article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
in the main text we have chosen a specific formalism to discuss electrodynamics in media .
this formalism encodes all the interactions of the medium with the electromagnetic radiation in the field @xmath100 through only the response function @xmath101 @xcite , without the need to introduce the macroscopic field @xmath102 . in this framework
we assess necessary or sufficient conditions for negative refraction . in this short section , building on the existing literature ,
we briefly explain why this is the appropriate formalism to undergo the first principle analysis we perform in the main text , the relation between this formalism and more well - known formalisms in term of the macroscopic fields @xmath100 and @xmath102 , the limit to the local formalism in term of the traditional functions @xmath85 and @xmath86 , and we comment on some subtleties .
the formalism for electrodynamics in media based on the response function @xmath101 is valid for any system @xcite in the linear response regime , without the need of any other specific assumption @xcite .
it encodes , as special case , the usual local formalism in term of the fields @xmath103 , @xmath104 , @xmath100 , @xmath102 and the electric permittivity @xmath0 and magnetic permeability @xmath2 , but it is much more general because it includes the spatial non - local effects disregarded in the local formalism @xcite .
the @xmath87 formalism indeed reduces to the local formalism in the limit of small spatial non - locality and slowly varying fields .
the possibility to write maxwell equations in media in term of a single additional macroscopic field @xmath100 , in addition to the the fields @xmath103 and @xmath104 , without introducing the field @xmath102 field is indeed quite well known @xcite . in the case of spatially homogeneous and isotropic invariant systems
the relation between the two formalisms can be easily made explicit @xcite .
indeed for spatial non - local media maxwell equations and the electromagnetic response of the medium can be equivalently written in term of the two vector fields @xmath105 , and @xmath102 , and the non - local scalar functions @xmath106 and @xmath107 : @xmath108 @xmath103 and @xmath109 @xmath102
. these two non - local formalisms are actually completely equivalent @xcite . for spatially homogeneous and isotropic media
the function @xmath101 depends only on two scalar functions : @xmath27 and @xmath28 , as explained in the main text .
the formalism in term of @xmath29 and @xmath100 , described in the main text , translates the fundamental fact that all the interactions between the electromagnetic field and the matter , including both electric and magnetic effects , are only due to the electromagnetic current @xmath110
. it indeed avoids to separate such effects in term of the two vectors : polarisation and the magnetisation , as instead done by the @xmath111 and @xmath112 formalism .
instead @xmath29 are the only fundamental functions : they satisfy all the properties of response functions @xcite , they are well defined for all values of @xmath15 and @xmath36 , and they are actually simply real linear functions of the causal correlator of electromagnetic currents : the fundamental observable encoding all the interaction between matter and electromagnetism .
such formalism is more appropriate and easier to handle in the case of time - varying fields . for time - varying fields the usual split of the current in term of polarisation and magnetisation vectors is indeed absolutely arbitrary @xcite .
the @xmath29 formalism implies to formally fix the magnetisation to zero and define the current only in term of the time derivative of the polarisation vector .
the consequence is that the magnetic effects are automatically encoded in the @xmath36 dependency of the dielectric tensor @xmath29 .
the relations between the two formalisms goes as follow @xcite : @xmath113 even if @xmath29 are the most fundamental objects from a theory perspective , it is nevertheless worth to state that the formalism in term of @xmath106 and @xmath107 results to be more adapted to describe static or almost static fields ( that could be more cumbersome in term of @xmath29 ) .
indeed maxwell equations allow to encode all the effects of a time varying magnetic field @xmath114 in the spatial variation of the transverse component of the electric field @xmath115 , in fourier : @xmath116 . from this equation
it is however clear that the limit of small spatial non - locality , i.e. @xmath117 , and slowly varying fields , i.e. @xmath118 , should be performed at the same time to obtain a non zero magnetic field .
this makes less straightforward to describe static magnetic fields in term of the @xmath29 , compared to the @xmath106 and @xmath107 formalism .
even if the formalism in term of @xmath106 and @xmath107 , could appear more familiar to the reader it has instead some subtleties that are rarely stated in literature , and it is worth discussing here @xcite . clarifying these subtleties would probably make clearer why the @xmath29 formalism should be preferred for a first principles analysis . as explained in the main text
, the causality of the system implies that @xmath29 satisfy the kramers - kronig relations , and the associated sum rules .
in particular the sum rule for the imaginary part of @xmath29 , discussed in the main text , and the second equation in ( [ eq8 ] ) , imply the equation : @xmath119 this should be valid for any value of @xmath36 , and it implies that the imaginary part of @xmath107 has to change sign in the interval @xmath120 $ ] .
this is in clear contrast with the case of @xmath29 , whose imaginary part is of definite sign for all frequencies ( and real positive @xmath36 ) , due to the passivity of the medium ( see the discussion in the main text ) .
such apparently wired fact is a direct consequence of the causality of the system .
both the formalism in term of @xmath29 , or the one in term of @xmath121 and @xmath122 , reduce to the usual local electrodynamics in term of the @xmath36 independent functions @xmath0 and @xmath2 in the limit of small spatial dispersion @xcite .
however such limit can somehow be subtle for what concern the magnetic response . indeed in the local limit : @xmath123 , while it can be shown that the function @xmath107 assumes the meaning of the magnetic permeability @xmath2 only in the narrow frequency band around @xmath95 for small value ok @xmath36 @xcite . in this limit
the resulting function satisfies all the usual properties of a response function @xcite , however such properties are in general not valid for the full set of frequencies . in particular it can be shown that the imaginary part of the function obtained taking naively the limit @xmath124 for @xmath122 should not have definite sign .
this is once again a consequence of the causality of the system .
nevertheless the resulting function is not interpretable as a magnetic permeability in that range of frequency .
it should hence be clear that a first principles analysis of the response of a generic medium can not be performed in the local formalism due to two main limitations .
the local formalism disregards spatial non - local effects that can become relevant for certain media or for certain range of parameters or frequencies .
for example some well - known corrections to the local formalism , due to spatial non - locality , come from the microscopic structure of the medium as in the case of crystals , for short wave length , or due to long range coherent phenomena as for example for plasmas , conducting or hydrodynamic media , where such corrections are relevant also at low frequencies .
the other limitation is that the function @xmath2 is a response function , and it actually describes the magnetic response of a medium , only for pretty limited set of frequencies @xcite around @xmath95 , and moreover it disguises in non - transparent way some weak spatial non - localities of the system . on the other hand in the spatially non - local formalism in term of @xmath29 these functions
are well defined response functions for all frequencies and wave vectors .
the local formalism results appropriate and better suited only for almost homogeneous and almost constant electric and magnetic fields @xcite .
in this section we briefly illustrate some explicit examples of spatial non - local electromagnetic responses relevant for phenomenological and experimental applications , and we discuss the effect of spatial non - locality on negative refraction for the associated media .
the aim of this analysis is to provide some explicit and concrete examples for the general theoretical analysis performed in the main text .
in particular we provide some analysis for the response function introduced in the main text ( see also equation ( [ eq29 ] ) here below ) in term of generic resonance , and in particular for the limits in which it describes well - known systems in literature .
a general analysis and classification of electromagnetic responses for spatially non - local media is left for future investigations .
we restrict our analysis to the transverse electric field and we consider the pretty general spatially non - local permittivity introduced in the main text : @xmath82 with @xmath83 , and @xmath84 real positive constants . as stated in the mail text , it can be easily check that ( [ eq29 ] ) is a good response function analytic in the complex upper half @xmath15 plane , and moreover its imaginary part is non - negative definite .
it represents the response of a spatially non - local medium near a resonance .
the function @xmath125 in ( [ eq29 ] ) reduces to two physically well understood cases for particular values of the parameters : the response of a charged viscous fluid ( like possibly electrons in certain metals @xcite ) , or the spatially non - local lorentz model used to describe the electromagnetic response of certain semiconductors near exciton - polariton resonance , as discussed for example in @xcite , see below .
using @xmath125 in ( [ eq29 ] ) we are able to illustrate some main points of the analysis performed in the main text . in particular we can see explicitly that while spatially non - locality is a necessary condition for negative refraction , it is instead in general not a sufficient condition ( contrary to what happens for the special case of a @xmath65 spatial non - local corrections to the local case , as discussed in the main text ) . here
below we provide examples of models where negative refraction is always present , models for which it is never present , and models where negative refraction appears only for certain values of the parameters of the medium , while it is absent for others .
the strategy we will use is to start from local models , in which the @xmath65 dependency of the function ( [ eq29 ] ) is not present or very weak , and slowly increase the parameters corresponding to spatial non - locality : i.e. @xmath126 and @xmath127 in ( [ eq29 ] ) . for the particular choice of parameters :
@xmath128 , @xmath129 , @xmath130 , and @xmath131 , the model ( [ eq29 ] ) reduces to the response function of a viscous charged fluids in the presence of dissipation @xcite : @xmath132 such function can be used to discuss for example the response to the electromagnetic field of electrons in certain metals . in this case
, the electrons correlation generates an effective viscosity , @xmath133 in equation ( [ eq30 ] ) , that hence induces the related spatial non - locality @xcite , while the lattice of positively charged atoms breaks translation invariance and it hence introduces an effective dissipation of the medium , represented by @xmath134 in equation ( [ eq30 ] ) .
this encodes the non - conservation of electrons momentum at microscopic scale , due for example to the presence of a microscopic lattice .
@xmath135 is the plasma frequency as discussed in the main text . in the particular case of @xmath136 ,
i.e. restoration of translation invariance at microscopic scale , the @xmath125 in ( [ eq30 ] ) reduces to the one of charged fluids in absence of impurities or lattice as it can be straightforward derived using navier - stokes plus maxwell equations . in this case
it is possible to show that the medium has always two refractive indices , one of which is always negative for frequency below @xmath135 @xcite .
we plot this result in figure [ fig1 ] : the refractive index associated to the dashed lines is indeed always negative . ) @xmath137 re(@xmath36 ) for the model in equation ( [ eq30 ] ) with @xmath138 .
continuos lines are associated to the first refractive index , while dashed lines are associated to the second refractive index .
the plots are done fixing @xmath139 and varying the values of the spatially non - local parameter @xmath133 .
three values of @xmath133 are represented , respectively @xmath140.,scaledwidth=40.0% ] for the particular choice of parameters : @xmath141 , @xmath142 , @xmath143 , @xmath144 , and @xmath145 , the model ( @xmath146 ) reduces to the response function of the lorentz model modified by spatial non - local effects parametrized by the coefficient @xmath147 : @xmath148 as shown in the figure [ fig2 ] this is a nice example of spatial non - local response with two refractive indices , both positive for all values of the parameters .
it is an explicit example showing that spatial non - locality is a necessary , but in general not a sufficient condition for negative refraction . )
@xmath137 re(@xmath36 ) for the model in equation ( [ eq31 ] ) .
continuos lines are associated to the first refractive index , while dashed lines are associated to the second refractive index .
the plots are done fixing @xmath149 , @xmath150 , and @xmath151 and varying the values of the spatially non - local parameter @xmath147 .
three values of @xmath147 are represented , respectively @xmath152 .
, scaledwidth=50.0% ] in this sub - section we show , as stated in the main text , that the general model represented in equation ( [ eq29 ] ) exhibits both negative and positive refraction depending on frequencies and the values of parameters . in figure [ fig3 ]
we indeed show that , varying the values of one of the spatial non - local parameters , we observe that there is a threshold of non - locality above which one of the two refractive indices becomes negative . )
@xmath137 re(@xmath36 ) for the model in equation ( [ eq29 ] ) .
continuos lines are associated to the first refractive index , while dashed lines are associated to the second refractive index .
the plots are done fixing @xmath153 , @xmath154 , @xmath155 , and @xmath156 and varying the values of the spatially non - local parameter @xmath127 .
three values of @xmath127 are represented , respectively @xmath157 .
, scaledwidth=50.0% ] v. g. veselago , sov .
usp . * 10 * , 509 ( 1968 ) .
d. r. smith , w. j. padilla , d. c. vier , s.c .
nemat - nasser , and s. schultz , phys .
lett . * 84 * , 4184 ( 2000 ) .
j. b. pendry , phys . rev
. lett . * 85 * , 3966 ( 2000 ) .
r. a. shelby , d. r. smith , s. schultz , science * 292 * , 77 ( 2001 ) .
veselago , v. , braginsky , l. , shklover , v. , and hafner , c. , j. comput .
. nanosci . * 3 * , 130 , ( 2006 ) .
a. v. chebykin , a. a. orlov , c. r. simovski , yu .
s. kivshar , and p. a. belov , phys .
b * 86 * , 115420 ( 2012 ) .
v. grigoriev , g. demesy , j. wenger , and n. bonod , phys .
b * 89 * , 245102 ( 2014 ) .
p. a. belov , r. marques , s. i. maslovski , i. s. nefedov , m. silveirinha , c. r. simovski , and s. a. tretyakov , phys .
b * 67 * , 113103 ( 2003 ) .
a. a. orlov , p. m. voroshilov , p. a. belov , and y. s. kivshar , phys .
b * 84 * , 045424 ( 2011 )
. a. v. chebykin , a. a. orlov , a. v. vozianova , s. i. maslovski , yu .
s. kivshar , and p. a. belov , phys .
b * 84 * , 115438 ( 2011 ) .
andrea al , phys .
b * 83 * , 081102(r ) ( 2011 ) .
d. forcella , j. zaanen , d. valentinis and d. van der marel , phys . rev .
b * 90 * , 035143 ( 2014 ) .
a. amariti , d. forcella and a. mariotti , jhep * 1301 * ( 2013 ) 105 .
m. i. stockman , phys .
* 98 * , 177404 ( 2007 ) . v. m. agranovich , y. n. gartstein , phys .
usp . * 49 * , 1029 ( 2006 ) . v. m. agranovich , y. n. gartstein , _ crystal optics with spatial dispersion , and excitons _
( springer - verlag berlin heidelberg gmbh 1984 , second corrected and updated edition . ) d. b. melrose and r. c. mcphedran , _ electromagnetic processes in dispersive media : a treatment on the dielectric tensor _ , ( cambridge : cambridge univ .
press , 1991 ) .
l. d. landau and e. m. lifshitz , _ electrodynamics of continuous media _ , ( pergamon press , 1960 ) .
a. a. rukhadze and v. p. silin , phys .
* 4 * , 459 ( 1961 ) . because we consider homogeneous invariant systems , the theory presented in the paper , and in particular eq .
( [ eq3 ] ) , describes only the bulk response of a medium , and not its boundaries properties .
l. p. kadanoff and p. c. martin , ann .
phys * 24 * , 419 ( 1963 ) .
d. forster , _ hydrodynamic fluctuations , broken symmetries and correlation functions _ , ( benjamin - cummings , reading 1975 ) .
for isotropic and non - gyrotropic systems the response functions depend only on @xmath160 , see equation ( [ eq12 ] ) .
m. altarelli , d. l. dexter , h. m. nussenzveig and d. y. smith , phys .
b * 6 * , 4502 ( 1972 ) .
it is important to stress that equation ( [ eq6 ] ) does not hold true for generic complex values of @xmath36 .
moreover for complex @xmath36 , i m @xmath161 does not have in general a thermodynamic interpretation as being proportional to the density of heat liberated in the medium per unit time , see for example @xcite .
this expansion is strictly valid only for small values of @xmath162 , while a similar expansion in term of @xmath163 should be considered for large values of @xmath162 , see for example @xcite .
hence our sufficient condition for negative refraction should be interpreted as strictly valid for small values of @xmath162 . in isotropic media
the group and phase velocities are both along the direction defined by @xmath16 .
we report only the coefficients of these vectors that define the versus of propagation . s. m. mikki and a. a. kishk , prog .
b * 14 * , 149 ( 2009 ) . for local media
this condition is equivalent to stating that the poynting vector points in the direction opposite to that of the phase velocity .
see for example m.w .
mccall , a. lakhtakia , and w.s .
weiglhofer , eur .
* 23 * , 353 ( 2002 ) .
r. a. depine and a. lakhtakia , microw .
. lett . * 41 * , 315 ( 2004 ) .
g. h. cocoletzi , and w. l. mochan , surf .
rep . * 57 * , 1 ( 2005 ) .
this result can be analyzed using a different approach based on sum rules for the refractive index .
it is not described here for the sake of consistency and brevity , and will be reported elsewhere .
markel , opt .
express * 16 * , 19152 ( 2008 ) . strictly speaking ,
this is true for spatially homogenous systems .
namely @xmath87 is able to describe the response of a generic medium in the bulk , while the presence of boundaries requires additional informations , such as for example : additional boundary conditions , or the introduction of additional information coming from a microscopic analysis of the medium .
when a macroscopic approach is possible the analysis should in general be performed in term of the response function @xmath164 in the direct space . | the importance of spatial non - locality in the description of negative refraction in electromagnetic materials has been put forward recently .
we develop a theory of negative refraction in homogeneous and isotropic media , based on first principles , and that includes non locality in its full generality .
the theory shows that both dissipation and spatial non locality are necessary conditions for the existence of negative refraction .
it also provides a sufficient condition in materials with weak spatial non locality .
these fundamental results should have broad implications in the theoretical and practical analyses of negative refraction of electromagnetic and other kinds of waves .
the study and design of artificial materials ( metamaterials ) with exotic electromagnetic properties has attracted a lot of attention both at the theoretical and experimental levels .
negative refraction has remained one of the most intriguing properties : in a certain range of frequencies , the energy flow is opposite to the direction of the phase velocity @xcite , offering a new potential for the design of components and devices ( the flat lens being the most famous example ) .
the investigations of the conditions for negative refraction have been originally carried out using the local description of electrodynamics in continuous media in terms of the electric permittivity ( dielectric function ) @xmath0 and magnetic permeability @xmath1 ( for a review see for example ref .
@xcite ) .
the first metamaterials exhibiting negative refraction were severely limited by absorption , and various strategies have been explored to design almost transparent media
. nevertheless , basic principles of electromagnetic waves propagation impose general constraints on the response functions of materials , and on the conditions for negative refraction .
for example in ref .
@xcite the principle of causality has been used to argue that , for spatially local materials , dissipation is necessary for the existence of negative refraction .
more recently , it has been pointed out that the use of local electrodynamics to analyse negative refraction in certain artificial materials has some intrinsic flaw .
the local effective @xmath0 and @xmath2 , usually obtained by fitting reflectivity or transmissivity curves , exhibit physical inconsistencies .
the necessity to include spatial non - local corrections has been put forward , in order to overcome the inconsistencies in the recovered parameters @xcite .
independently , it has been shown that spatial non - locality is the fundamental ingredient for the existence of negative refraction in some specific materials ( including natural materials ) @xcite .
although the role of spatial non locality in the phenomenon of negative refraction has been highlighted in recent studies , a theory providing a general understanding of the conditions for negative refraction , based on first principles and including spatial non locality , is still missing .
the purpose of this letter is to develop such a theory , and to analyse its consequences in terms of necessary and sufficient conditions for the observation of negative refraction in natural and artificial electromagnetic materials .
the analysis leads to the intriguing conclusion that negative refraction for electromagnetic waves is not possible in absence of spatial non - locality and dissipation .
it also gives some insight into a sufficient condition to observe negative refraction in media with weak non locality .
these results provide a sound basis both for the theoretical investigation of negative refraction in wave physics , and for the design and analysis of real materials .
maxwell s equations for electrodynamics in material media can be written in full generality in the following form @xcite : @xmath3 where @xmath4 and @xmath5 are external current and charge densities and @xmath6 is the vacuum permeability . in the framework of linear response theory , the field @xmath7 is defined as @xcite @xmath8 with @xmath9 the volume of the medium , @xmath10 the vacuum permittivity , and where implicit summation over repeated indices is assumed .
the dielectric function @xmath11 is proportional to the retarded correlator of the current density , and encodes both the electric and magnetic response of the medium @xcite . in particular , in this formalism , there is no need to introduce an additional field @xmath12 , provided that the full dispersive and non - local behaviour of the dielectric function is taken into account ( see the appendix ) . for simplicity
, we restrict the analysis to linear , isotropic , non - gyrotropic and homogenous materials , and leave generalisations for further studies . apart from this restriction , all results are derived from first principles , without any specification of a particular medium @xcite . for translation invariant media , one has @xmath13 , and in fourier space eq .
reads as @xmath14 where @xmath15 is the frequency and @xmath16 is the wave vector that describes spatial non - locality .
we stress that all magnetic effects , such as the magnetic response usually described in term of a local permeability @xmath2 , or even the magnetoelectric activity , can be encoded into specific spatial non - local properties of the dielectric function @xcite .
a discussion on the relation between this formalism and the usual approach in terms of the fields @xmath17 and @xmath12 is provided in the appendix , as well as a justification of the relevance of this formalism for a first principles analysis of the electromagnetic response of material media .
the dielectric function @xmath18 has to satisfy various properties @xcite .
causality implies that @xmath19 vanishes for @xmath20 . in the frequency domain
, this means that @xmath18 is an analytic function of @xmath15 in the positive half - part of the complex plane for which @xmath21 .
this property leads to the kramers - kronig relations @xmath22 where @xmath23 stands for principal value .
it is important to highlight that the kramers - kronig relations are valid for all values of @xmath16 , that essentially plays the role of a parameter in the equations , and is , at this stage , independent on @xmath15 . in isotropic media ,
the dielectric function takes the form @xmath24 where @xmath25 are the components of @xmath16 and @xmath26 @xcite .
the response of the material is expressed in term of two scalar functions only , the transverse and the longitudinal dielectric functions @xmath27 and @xmath28 , that both satisfy the kramers - kronig relations . from the kramers - kronig relations and the asymptotic behaviour of @xmath29 when @xmath30 , various sum rules can be derived @xcite , that encode important information and constraints on @xmath29 . an important sum rule that will be very useful in the analysis
is @xcite @xmath31 where @xmath32 is the square of the plasma frequency , @xmath33 being the electron density of the material , and @xmath34 and @xmath35 the charge and the mass of the electron , respectively .
note that eq . is valid for all values of @xmath36 .
finally , for any passive medium , and for real positive values of @xmath36 and @xmath15 @xcite , the dielectric function satisfies @xcite @xmath37 equations ( [ eq4 ] ) , ( [ eq5 ] ) and ( [ eq6 ] ) are fundamental relations that constrain the actual electrodynamic properties of any causal and passive medium .
the frequency @xmath15 and the wave vector @xmath36 have been treated so far as independent variables , _
i.e. _ , the system is considered `` off - shell '' , as usual in linear response theory @xcite ( due to the presence of external sources , @xmath15 and @xmath36 do not have to satisfy any relation @xcite ) .
however when considering the propagation of electromagnetic waves in a medium without external charges or currents , maxwell s equations impose the dispersion relations @xmath38 for the transverse and longitudinal components of the electric field @xmath39 , respectively , with @xmath40 the speed of light in vacuum .
equations define a set of relations between the wave vector @xmath36 and the frequency @xmath15 .
these relations provide further restrictions to the electrodynamics in the medium . in the presence of spatial non - locality , both the transverse and longitudinal components of the electric field
can propagate . in this letter
we focus on the response to transverse electromagnetic waves .
we shall first discuss the particular case of transparent media . in this case
, negative refraction means that , in a certain range of frequencies , phase and group velocities have opposite directions . indeed , both the electromagnetic energy density and the poynting vector are well defined , the latter being simply the product of the electromagnetic energy density and the group velocity ( the group velocity equals the energy velocity , at least for media with weak spatial non - locality @xcite ) . for local media
, it has already been argued that negative refraction can not be observed in absence of dissipation @xcite .
we will now show that the same conclusion holds in the presence of spatial non - locality , namely , that for transverse waves , negative refraction is not possible in the frequency region of transparency of a medium . in the transparency region
one has @xmath41 , and the kramers - kronig relations involve standard integrations . deriving eq . with respect to frequency @xcite , and using eq .
, we obtain @xmath42 } { d\omega } \geq 0\ ] ] for all frequencies in the transparency region ( this is not the case in regions in which absorption can not be disregarded , where the dielectric function can satisfy @xmath43/d\omega < 0 $ ] , known as anomalous dispersion ) . for transverse waves eq
. implies @xmath44=k^2c^2/\omega^2=\epsilon_{t}[\omega , k(\omega)]$ ] , with @xmath45 the transverse refractive index , and hence the dispersion relation @xmath46 .
the transverse group velocity is defined as @xmath47^{-1}= c [ n_t+\omega d n_t / d \omega]^{-1}$ ] , while the phase velocity is @xmath48 @xcite . using eq . ,
noticing that @xmath36 and @xmath45 are real functions of frequency , we obtain @xmath49 , or equivalently @xmath50 this implies that , as a consequence of causality , negative refraction for transverse waves is not possible in the transparency spectral region of a medium .
we shall now perform a general analysis of the conditions for negative refraction , including both dissipation and spatial non - locality . for non - transparent media , a consistent definition of concepts such as energy density , energy flux , or
heat dissipation rate is out of reach ( see for example ref .
@xcite ) .
in particular the group velocity does not define the direction of the energy flow .
we have to find an alternative way to assess the direction of the energy flow based on the wave vector . to proceed , we use eq .
( [ eq7 ] ) to obtain the dispersion relation @xmath46 .
since we consider passive media , the direction of the energy flux is dictated by the direction of dissipation of energy , or equivalently of the damping of the wave amplitude , the latter being given by the sign of @xmath51 .
the direction of the phase velocity is given by @xmath52 .
as a result , a general condition for negative refraction is the existence of frequencies @xmath15 such that @xcite @xmath53\cdot \hbox{im } [ k(\omega ) ] < 0 \ , \ ] ] meaning that the damping of the wave amplitude occurs in a direction opposite to the phase velocity .
we have demonstrated previously that for both local and non - local media , negative refraction is not possible without dissipation .
we shall now show that spatial non - locality itself is a necessary condition for negative refraction .
let us consider a dissipative medium for which spatial non - locality is assumed to be negligible . for transverse waves
, we have @xmath54 and the dispersion relation eq .
simplifies into @xmath55 for convenience we choose to represent any complex number @xmath56 as @xmath57 , with @xmath58 and @xmath59 . due to
eq . ,
@xmath60 for real frequencies , and the phase @xmath61 must satisfy @xmath62 .
the phase of its square root then satisfies @xmath63 , and hence @xmath64 equation is clearly in contradiction with condition , and we are left with the conclusion that spatial non - locality and dissipation are necessary conditions for the existence of negative refraction .
this is the first important result in this letter .
although a related discussion can be found in ref .
@xcite , these necessary conditions have not been stated previously to our knowledge .
finally , let us note that the necessity of spatial non - locality also holds for longitudinal waves . indeed , in absence of spatial non - locality the longitudinal component of the electric field can not propagate , and hence can not exhibit negative refraction .
we now investigate the existence of a sufficient condition for negative refraction . for non - gyrotropic media , due to invariance under point
reflection , the dielectric functions @xmath29 can be expanded in powers of @xmath65 in the form ( contributions with odd powers of @xmath36 are forbidden by symmetry ) @xmath66 where @xmath67 denotes the largest intrinsic characteristic length of the material ( _ e.g. _ the lattice constant for crystals , the mean free path of conduction electrons for metals , the debye screening length for plasmas , etc ) , and vanishes in local media . inserting eq . into the sum rule , and imposing the validity of the resulting equation for any value of @xmath36 , we obtain the following set of equations : @xmath68 equations ( [ eq13 ] ) implies that for @xmath69 , the functions @xmath70 have to change sign in the interval @xmath71 $ ] .
this apparently surprising result is a direct consequence of causality ( for further details see the appendix ) .
also note that it is not in contradiction with eq . since it only applies separately for each @xmath72 with @xmath73 , and i m @xmath74 , while eq .
implies that the imaginary part of the total sum ( [ eq12 ] ) ( for real @xmath36 ) is positive . for transverse waves , and for isotropic and non - gyrotopic media
, we shall now prove that the first non - local correction to the dielectric function gives a sufficient condition for the existence of negative refraction .
let us consider the dielectric function @xcite @xmath75 inserting eq . into
, we obtain @xmath76 where @xmath77^{-1}$ ] . from eq .
we have deduced that there exist frequencies @xmath15 such that @xmath78 , leading to @xmath79 . since @xmath80 , we can state that @xmath81 this condition on the phase of the vector amounts to showing that frequencies for which eq .
( defining negative refraction ) is satisfied necessarily exist .
we conclude that a non - local medium , that can be represented as the most general first order correction ( allowed by symmetry ) to a local response eq .
, necessarily has a range of frequencies in which negative refraction is observed .
this is the second important result in this letter .
the existence of the necessary condition based on a generic @xmath65 correction to the local response model deserves two comments .
first , applying the same analysis to longitudinal waves does not allow us to conclude on the existence of a sufficient condition for negative refraction in this case .
second , we stress that while dissipation and spatial non - locality are necessary conditions , they are not in general sufficient conditions for negative refraction . for example , let us consider a non - local transverse dielectric function of the form @xmath82 representing the generic response of a spatially non - local ( isotropic , homogeneous and passive ) medium near a resonance , @xmath83 and @xmath84 being real positive constants . for particular choices of the parameters
, it describes the response of charged viscous fluids ( _ e.g. _ electrons in certain metals @xcite ) , or the spatially non - local lorentz model ( _ e.g. _ exciton - polariton resonance for certain semiconductors @xcite ) .
changing the values of the coefficients of the @xmath65 terms , one can switch from a negative refraction regime to a positive refraction regime at low frequencies , as previously shown for the electrons hydrodynamic model @xcite ( more details and other examples are given in the appendix ) .
the analysis therefore leads to the conclusion that spatial non - locality supports , but does not necessarily imply negative refraction @xcite . before concluding it is worth commenting on the relation between the formalism used in this letter and the more familiar local formalism
the usual local analysis , in term of the dielectric function @xmath85 and a magnetic permeability @xmath86 is clearly included in the non - local formalism described in this letter in term of a single spatially non - local response function @xmath87 .
it indeed emerges as a particular case in the limit @xmath88 and @xmath89 ( see refs .
@xcite and the appendix ) . in this limit
, our result that spatial non - locality is a necessary condition for negative refraction translates into the well known fact that , in the local limit , to observe negative refraction it is necessary to have both @xmath90 and @xmath91 . the sufficient condition for negative refraction , usually stated in the local formalism in the form @xmath92 \ , |\mu(\omega)|+$ ]
@xmath93 \ , |\epsilon(\omega)| < 0 $ ] ( see _ e.g. _ refs .
@xcite ) is consistent with ( although not implied by ) our analysis .
indeed , the sufficient condition of a @xmath65 non - local correction does not require any constraints on the general form of the non - local permittivity eq .
( [ eq16 ] ) .
however , one should not conclude that the theory described in this letter is a mere generalisation of a well - established approach .
the use , in the first place , of a non - local description of the electrodynamic response is fundamental at least for three reasons .
first , it is important to note that the meaning of @xmath86 as a true response function [ with @xmath94 for passive media ] is only valid for very small @xmath36 _ and _ in a limited range of frequencies in the neighbourhood @xmath95 @xcite , see also the appendix . outside this range ,
@xmath86 looses its meaning of a magnetic permeability , and its imaginary part has in general an indefinite sign ( see the appendix ) .
this is actually the reason behind the sufficient condition for negative refraction proven in this letter [ the function @xmath96 in eq .
can be understood as an effective permeability ] .
second , the analysis based on a non - local dielectric function reveals that @xmath86 , even if apparently a local response , encodes ( some of ) the non - local effects to order @xmath65 initially included in the non - local dielectric function @xmath97 ( see the appendix for a proof ) .
therefore , the existence of a permeability @xmath98 , as a condition for negative refraction in the local approach , can be seen as a disguised residual of spatial non - locality .
third , the local formalism , although appropriate for a phenomenological analysis of certain materials , can not be used for a first principles analysis of negative refraction on the full spectrum of frequencies , because of its limited range of validity around @xmath99 and @xmath89 @xcite . in the present letter we applied constraints coming essentially from causality [ except for eq .
( [ eq7 ] ) ] . in the local limit it is possible to include thermodynamical constraints as well , that are instead in general not valid in the non - local theory . in ref .
@xcite it has been indeed argued that thermodynamics forbids negative refraction in the local limit .
the addition of thermodynamical considerations hence strengthen once more the need of non - local approach , and the fact that negative refraction is realised exactly outside the regime of applicability of the @xmath85 and @xmath86 formalism . in conclusion , we have demonstrated that negative refraction of electromagnetic waves is intrinsically related to both dissipation and spatial non - locality .
their simultaneous presence is a necessary condition for negative refraction .
the implications of the theory developed in this letter are manyfold . at the fundamental level
, the analysis can be extended non - isotropic or non - homogeneous media , or to other kind of waves such as acoustic waves , for which the design of negative refraction materials is an active field as well . at a practical level ,
the results in this letter show that non locality should not be considered as a refinement of the existing theories , allowing one to improve in an incremental way the precision in the determination of the effective response functions of real materials .
instead , any inverse procedure should include in the first place non locality , since its exclusion is fundamentally inconsistent , and can lead to uncontrollable systematic errors .
finally , our results open the way to a systematic classification of media with or without negative refraction , providing a clear connection between this phenomenon and the dynamical properties of materials , as already shown in the case of metals @xcite .
d. f. would like to thank v. markel for relevant discussions .
this work was supported by labex wifi ( laboratory of excellence within the french program `` investments for the future '' ) under references anr-10-labx-24 and anr-10-idex-0001 - 02 psl*. |
the purpose of this appendix is to offer a proof of the following claim in the paper : _ the squared speed of the evolution of a quantum state generated by a hamiltonian @xmath98 is bounded above by twice the hilbert - schmidt norm @xmath99 of the hamiltonian , and the bound is attained if @xmath98 is horizontal .
_ we begin by establishing some properties of the space of horizontal hamiltonians .
recall that a hamiltonian @xmath98 is horizontal with respect to some state @xmath39 if and only if @xmath100 for all hamiltonians @xmath101 that leave @xmath102 invariant . for simplicity
, we fix the state @xmath102 to be the one whose homogeneous coordinates are given by @xmath103 .
it should be stressed , however , that the conclusions of the discussion below remain valid for any choice of @xmath39 , owing to the homogeneous nature of the state space @xmath104 .
let us write @xmath105 for the subgroup of su@xmath43 that leaves @xmath102 invariant and @xmath106 for the set of the corresponding generators ( i.e. its lie algebra ) .
then @xmath105 consists of all unit determinant matrices of the form @xmath107 and @xmath108 consists of all trace - free matrices of the form @xmath109 the orthogonal complement @xmath110 of this space with respect to the hilbert - schmidt norm coincides , up to a factor of @xmath111 , with the space of horizontal hamiltonians .
specifically , the set @xmath110 can be found by requiring that @xmath112 for all choices of @xmath113 and @xmath114 with @xmath115 .
here we have parameterised the full lie algebra @xmath116 using an imaginary @xmath117 , a vector @xmath118 and @xmath119 with @xmath120 .
it follows that @xmath110 consists of all matrices of the form @xmath121 where @xmath118 . as a consistency check
, we note that this space has @xmath61 real dimensions , which of course is the same as the dimensionality of the state space @xmath122 .
it follows that if @xmath98 is a horizontal hamiltonian , then @xmath123 on the other hand , if @xmath39 evolves according to the schrdinger equation , then the squared speed ( in the fubini - study metric ) of its evolution is given , on account of the anandan - aharonov relation , by four times the variance of @xmath98 : @xmath124 this establishes part of our claim : namely , by comparing and we see that the square of the evolution speed equals twice the hilbert - schmidt norm @xmath99 if @xmath98 is horizontal . to complete the proof ,
note that any hamiltonian @xmath98 can be decomposed into vertical and horizontal parts , i.e. it can be written in the form @xmath125 then @xmath126 but the trace on the right hand side is strictly negative if the vertical part of @xmath98 is nonzero .
therefore , the squared speed of evolution on state space , given by @xmath127 if the hamiltonian contains a vertical component .
this completes the proof of the claim . | the quantum navigation problem of finding the time - optimal control hamiltonian that transports a given initial state to a target state through quantum wind , that is , under the influence of external fields or potentials , is analysed . by lifting the problem from the state space to the space of unitary gates realising the required task , we are able to deduce the form of the solution to the problem by deriving a universal quantum speed limit .
the expression thus obtained indicates that further simplifications of this apparently difficult problem are possible if we switch to the interaction picture of quantum mechanics .
a complete solution to the navigation problem for an arbitrary quantum system is then obtained , and the behaviour of the solution is illustrated in the case of a two - level system . with the advances in the implementation of quantum technologies , the theoretical understanding of controlled quantum dynamics and , in particular , of their limits , is becoming increasingly important .
one aspect of such limits that has been investigated extensively in the literature concerns the time - optimal manoeuvring of quantum states @xcite .
if the time - evolution is unconstrained ( apart from a bound on the energy resource ) , then this amounts to finding the time - independent hamiltonian that generates maximum speed of evolution .
however , in general there can be a range of constraints that prohibits the implementation of such an elementary protocol , and various optimisations will have to be applied to determine time - dependent hamiltonians that generate the dynamics achieving required tasks .
an important class of problems arising in this context is the identification of the time - optimal quantum evolution under the influence of external fields or potentials that can not be easily eliminated in a laboratory .
solutions to such problems are indeed relevant to practical implementations of time - optimal controlled quantum dynamics because in real laboratories external influences ( e.g , electromagnetic fields ) are typically present
. problems of this kind can be thought of as representing the quantum counterpart of the classical zermelo navigation problem of finding the time - optimal control that takes a ship from one location to another , under the influence of external wind or currents @xcite . within the context of quantum zermelo problems , there are two distinct questions that arise , namely , finding the time - optimal hamiltonian ( i ) that generates a required unitary gate ; and ( ii ) that transports a given initial state to a required target state . in the context of constructing a unitary gate ,
this problem was formulated in @xcite , and solved more recently in @xcite .
the construction of the time - optimal hamiltonian that transports a given initial state to a target state , on the other hand , is _ a priori _ more challenging ( cf .
@xcite ) , because ( as shown below ) this involves a variational problem with free boundaries .
the purpose of the present paper is to derive the full solution to this latter problem for an arbitrary quantum system .
one way of addressing the quantum - state navigation problem is to work in the space of rays through the origin of the underlying hilbert space ( i.e. the complex projective space , or simply the ` state space ' ) .
then the schrdinger evolution generated by the background field gives rise to a hamiltonian vector field on the state space , and the task at hand can be formulated as a hamiltonian control problem , the solution of which is typically difficult to obtain .
we are nevertheless able to go forward by exploiting certain subtle geometric structures of the unitary group , which in turn allows us to efficiently make use of the results obtained in @xcite .
a remarkable feature of the solution that we obtain is that it is strikingly reminiscent of the analysis of quantum dynamics in the interaction - picture of schwinger and tomonaga @xcite . on account of this observation
we are able to revisit our analysis by employing a physical symmetry argument , rather than purely variational reasoning . as a consequence ,
a considerably simpler derivation of the solution emerges .
in fact , it turns out that by invoking the line of thinking behind the schwinger - tomonaga theory , a straightforward solution can even be found of closely - related and apparently difficult problems in geometry @xcite . for an illustration
we conclude the paper by presenting examples in the case of a two - level system .
let us begin by stating more explicitly the zermelo navigation problem for quantum states : for a given initial state @xmath0 and a final target state @xmath1 , the task of an experimentalist is to find the control hamiltonian @xmath2 such that the total hamiltonian @xmath3 will generate the transformation @xmath4 in the shortest possible time . here
@xmath5 represents the hamiltonian of the background field or potential that can not be manipulated .
since the objective is to realise a time - optimal transformation ( navigation ) , it is assumed that the available energy resource , which evidently has to be bounded , will be consumed fully .
this translates into the ` full throttle ' condition that the squared magnitude of the evolution speed generated by the control hamiltonian @xmath6 , which is given on account of the anandan - aharonov relation @xcite by four times the variance @xmath7 of @xmath6 , is held fixed at all times at the maximum attainable value .
( colour online ) _ unit circle in the wind_. in the absence of the wind , the ` unit - time ' circle is centred at the origin @xmath8 .
however , with the prevailing wind , with strength @xmath9 , the unit - time circle reachable from the origin is shifted in the direction of the wind .
since it takes one unit of time to reach @xmath10 , reaching the target @xmath11 will take @xmath12 units of time . ] to illustrate the nature of the task involved , let us first consider a more elementary problem of zermelo navigation on the plane @xmath13 , with constant ( in space and in time ) wind , represented by the vector @xmath14 .
suppose that , departing from the origin @xmath8 the desired destination is given by the endpoint of a vector @xmath15 .
to work out the time it takes to reach the destination , consider a circle whose radius is determined by the distance that can be reached with full speed over one unit of time . in the absence of wind the circle is clearly centred at the origin .
however , under the influence of the wind the centre of the circle is shifted by @xmath14 . for the moment
let us assume that the wind is not dominant , i.e. @xmath9 so that the origin remains inside the shifted circle .
this configuration is schematically illustrated in figure [ fig:1 ] . under the influence of the wind , therefore , to reach the endpoint @xmath11 of the vector @xmath15 , it suffices to determine the distance @xmath16 , where @xmath10 is the point of intersection of the vector @xmath17 with the shifted circle .
this follows on account of the fact that since it takes one unit of time to reach @xmath10 , it will take @xmath18 is given by @xmath19 where @xmath20 is the euclidean inner product , @xmath21 , and @xmath22 .
the foregoing analysis on @xmath13 extends naturally to the general case of navigation on a riemannian manifold of arbitrary dimension , albeit the euclidean inner product @xmath23 has to be replaced by the riemannian one .
this is because the total journey time can be decomposed into a large number of infinitesimal journey times , each on a tangent space of the manifold . in the case of optimal quantum navigation
, we recall that the space of pure states is equipped with the unitary - invariant fubini - study metric @xcite . the wind @xmath14 is then replaced by the hamiltonian symplectic flow generated by @xmath5 , whereas @xmath15 is some vector tangent to quantum state space .
the total journey time @xmath24 along a path on state space is given by the integral of @xmath25 , where @xmath15 is the velocity vector .
in particular , for the zermelo problem , @xmath15 is generated by @xmath3 , and this implies that @xmath25 in the quantum context is given by a function of the variance @xmath26 of @xmath5 , the variance @xmath27 of @xmath6 , and the covariance @xmath28 of the two hamiltonians .
the optimal control is obtained by minimising @xmath24 over all @xmath2 , subject to the full throttle constraint @xmath29 .
the no - dominance condition on the wind is required in the case of a navigation problem on an open manifold such as @xmath30 , since otherwise the target may never be reached . in the quantum context we are able to relax the requirement @xmath31 , since the manifold of pure states is compact , which means that if the wind against the shortest path is too strong , one can always go ` the other way around ' to reach the target state ( note incidentally that @xmath25 remains well defined in the limit @xmath32 since @xmath33 ) .
the hamiltonian control problem specified above is not straightforward to solve directly , and this leads us to employ an alternative approach based on the idea of lifting the problem from state space to the unitary group acting on the states .
in fact , from a physical point of view this is more appealing , since , although we are seeking the time - optimal path on state space , physical implementations are always carried out by constructing the hamiltonian that generates the desired unitary operator .
bunt3.pdf ( 72.5 , 79.5)@xmath34 ( 16.40 , 14 ) @xmath0 ( 19.50 , 58 ) @xmath35 ( 22.50 , 73 ) @xmath5 ( 68.50 , 22 ) @xmath1 ( 74.0 , 9.0 ) @xmath36 ( 76.0 , 52 ) @xmath37 ( 74.5 , 41.50 ) @xmath38 with this in mind , we proceed by considering a fibre space above the state space with the property that for any state @xmath39 , the fibre above it consists of elements @xmath40 of the special unitary group fulfilling the condition that @xmath41 . although our solution is applicable to both finite and infinite dimensional hilbert spaces , to simplify the discussion let us assume that the hilbert space is of finite dimension @xmath42 .
the totality of such fibres , when bundled together , then forms the group su@xmath43 , which acts on the state space ( the complex projective space @xmath36 ) .
this configuration is illustrated in figure [ fig:2 ] .
let @xmath34 be an element of the fibre above the target state @xmath1 , i.e. @xmath44 . then the problem of finding the time - optimal transformation @xmath45 translates into the problem of constructing the unitary gate @xmath34 from the identity element @xmath46 in the shortest possible time , under the influence of an external field .
this translation of the problem helps us because the optimal zermelo navigation for a specified unitary gate has been worked out recently @xcite , and we can make use of the result in deducing the optimal control hamiltonian for the state transfer ; however , we encounter two difficulties : ( a ) the ` target ' unitary gate @xmath34 is not unique since _ a priori _ it can be any element of the fibre above @xmath1 . in other words , if @xmath47 is any unitary operator that leaves the state @xmath1 invariant , then the unitary gate @xmath48 also satisfies the required condition @xmath49 so we have to deal with a free final boundary ; ( b ) the ` full throttle ' condition @xmath29 on state space is concerned with the evolution speed , whereas the condition used in @xcite for unitary gates appears to be distinct since it concerns the hilbert - schmidt norm @xmath50 . in what follows
we shall show that these two apparent issues evaporate once the relevant ` horizontality condition ' is imposed . to see this we begin by noting that since latexmath:[${\hat e}|\psi_i\rangle= the fibre above the initial state @xmath0 .
in fact ,
the fibre above @xmath0 coincides with the totality of unitary group elements @xmath52 that leave the initial state @xmath0 invariant .
intuitively , to reach from @xmath35 a target gate @xmath34 above @xmath1 in a timely manner we would like to move away from @xmath35 as quickly as possible , and this is achieved by manoeuvring _ horizontally _ ,
i.e. towards a direction that is orthogonal to the fibre .
that is to say , the choice of the initial control hamiltonian @xmath53 has to be such that @xmath54 for all hamiltonians @xmath55 that leave the initial state invariant ( i.e. @xmath53 has to be orthogonal to all generators of @xmath56 ) .
this suggests that the horizontality condition @xmath57 gives rise to the maximum evolution speed . to show this more explicitly we shall derive a kind of universal quantum speed limit .
the statement of the result we shall establish is as follows : the squared speed of the evolution of a quantum state generated by a hamiltonian @xmath58 , as defined by anandan and aharonov @xcite , is bounded above by twice the hilbert - schmidt norm @xmath59 of the hamiltonian , and the bound is attained if @xmath58 is horizontal .
the implication of this result is that under the horizontality requirement the norm condition and the maximum speed condition are equivalent , and this resolves the issue ( b ) raised above . as a consequence , we are able to borrow the results of @xcite to deduce that the time - optimal control hamiltonian @xmath2 is necessarily of the form @xmath60 notice that the horizontality condition is preserved under the adjoint action ( [ eq:2 ] ) , which is indeed required since the full - throttle condition has to be maintained throughout the operation the horizontality condition alone , of course , does not fix @xmath53 uniquely , since the constraint @xmath57 on @xmath53 leaves @xmath61 degrees of freedom . however , by demanding that the required gate should generate the target state @xmath1 , which imposes @xmath61 conditions , we are able to select a unique initial control hamiltonian @xmath53 . in other words , the freedom in the boundary is eliminated by the horizontality condition @xmath57 along with the boundary condition @xmath44 , and this resolves the issue ( a ) raised above .
before we turn to determining the initial control hamiltonian @xmath53 , and the amount of time @xmath24 that is needed to reach the target state under the optimal control , let us discuss the claim above on the universal speed limit ( cf .
@xcite ) . from a physical point of view
the claim is plausible if we make note of the following observation .
suppose that the hamiltonian contains a vertical component tangent to the fibre that leaves the state below invariant .
in this case , although energy is scarce , the experimentalist will be consuming energy that produces no work ( except for shifting the overall phase ) .
this is clearly not energy efficient .
it follows that the optimal performance is ensured by eliminating vertical components altogether . a more precise derivation , which essentially follows from a standard result on complex projective spaces outlined , e.g. , in kobayashi & nomizu @xcite ( see xi.10 ) , is given in the supplementary information .
having established the ` geodesic ' curve ( [ eq:2 ] ) that minimises the action @xmath62 we are now in the position to identify the initial control @xmath53 so that the target state is reached . to this end
we observe that the form ( [ eq:2 ] ) of the solution is indicative of the analysis in the interaction picture of quantum mechanics , if we view @xmath5 as the ` free ' and @xmath53 as the ` interaction ' hamiltonian .
alternatively stated , if we change from the ` rest frame ' to the ` moving frame ' on the state space according to the wind @xmath5 , then without the control hamiltonian the initial state remains fixed , while the target state moves against the wind according to @xmath63 .
evidently , the state @xmath64 thus defined is the solution to the ( nonrelativistic version of ) the tomonaga - schwinger equation .
in particular , standard results in the interaction - picture analysis @xcite show that the time - ordered product of the unitary group generated by the hamiltonian @xmath65 gives rise to the evolution equation of the form @xmath66 this result can also be verified directly ( cf .
@xcite ) by differentiation : writing @xmath67 , we find @xmath68 .
as indicated above , to fix the initial condition @xmath53 we need to make use of the boundary condition , which implies that @xmath69 must hold for some @xmath24 that is to be determined .
observe that there are @xmath61 unknowns in @xmath53 , and we have an additional unknown parameter @xmath24 . the boundary condition ( [ eq:4 ] )
gives rise to @xmath61 constraints .
together with the norm condition @xmath50 we are thus able to fix all the unknowns , and this in turn solves the quantum - state zermelo navigation problem .
flows3.pdf ( 3 , 54 ) ( 50.3 , 75)@xmath70 ( 53.8 , 74 ) * . *
( 40.5 , 14 ) @xmath71 ( 44.7 , 13.3 ) * .
* to proceed with this let us rewrite ( [ eq:4 ] ) in the form @xmath72 although the transformation from ( [ eq:4 ] ) to ( [ eq:5 ] ) is in itself trivial , it sheds a different light on the problem at hand . as indicated above ,
if we transform to the moving frame generated by the wind such that the initial state remains stationary , then the objective becomes ` hitting ' a moving target @xmath73 as quickly as possible .
exactly how long it will take to achieve this task depends on how fast the motion towards the target can be , but for a fixed speed the shortest time @xmath24 achievable is attained by following the geodesic path on the state space joining @xmath0 and @xmath74 . on the other hand ,
the curve @xmath75 joining @xmath0 and @xmath76 is a geodesic on the fubini - study manifold if and only if @xmath53 is horizontal , and this is indeed the situation we have in ( [ eq:5 ] ) . to determine @xmath53 explicitly ,
let @xmath77 be the angular distance between @xmath0 and @xmath74 defined by the relation : @xmath78 evidently , the angular distance @xmath79 is given by the speed @xmath80 of the evolution generated by @xmath53 multiplied by the duration @xmath24 of the journey , but the evolution speed is fixed on account of the condition @xmath50 to @xmath81 , so we deduce that @xmath82 since @xmath5 , @xmath0 , and @xmath1 are given , we see therefore that the smallest positive root of ( [ eq:7 ] ) uniquely determines @xmath24 .
once @xmath24 is fixed , the optimal control hamiltonian at time zero can be obtained explicitly by making use of the result in @xcite : @xmath83 note that without loss of generality we can adjust the overall phase such that @xmath84 ) , one can easily verify that this hamiltonian is indeed horizontal .
this completes the derivation of the solution to the quantum - state zermelo navigation problem : the optimal control hamiltonian is given by ( [ eq:2 ] ) , where @xmath53 is given by ( [ eq:8 ] ) and where the parameter @xmath24 is determined by ( [ eq:7 ] ) . to gain further intuition on the behaviour of the solution , in figure [ fig:3 ] we plot the trajectories of the quantum state generated by the hamiltonian @xmath85 in the case of a two - level system .
in particular , we show the time - optimal paths for a range of wind strengths , indicating for instance that if the head wind is too strong , then the optimal strategy is to switch the direction of the manoeuvre to turn the head wind into a tailwind . to understand the minimum journey times associated with different wind strengths we have also considered a one - parameter family of winds @xmath86 , corresponding to the setup shown in figure [ fig:3 ] , and determined the optimal time @xmath87 for a range of @xmath88 ; the result is shown in figure [ fig:4 ] .
( colour online ) _ journey time on the bloch sphere_. the trajectories @xmath89 sketched in figure [ fig:3 ] correspond to the wind hamiltonian @xmath86 for @xmath90 , for the choice of @xmath5 sketched therein . here
, journey time @xmath87 is plotted for a continuous range of @xmath88 , showing that once the wind strength is sufficiently strong , by going the other way around , @xmath24 can be reduced by turning head wind into a tailwind . ] in summary , we have been able to deduce in closed form the solution to the quantum zermelo problem in complete generality .
since the result obtained takes a rather simple form , we expect that a practical implementation is feasible in a number of realistic situations .
of course , for some applications the controllable degrees of freedom in the hamiltonian can be limited , in which case additional constraints will have to be imposed in finding the optimal control some steps towards this direction in the context of unitary gates have been initiated in @xcite . in fact , further progress can again be made by switching to the interaction picture : if we write @xmath91 for the @xmath92 constraints on the control hamiltonian , then in the moving frame the constraints will take for each @xmath93 the form @xmath94 , where @xmath95 is the generator of the dynamics @xmath96 .
the analysis of @xcite , for instance , can then be applied with essentially just one modification , namely , that the target state is moving .
the initial control hamiltonian can then be identified by adapting the strategy proposed in @xcite .
it is worth noting that the use of the interaction picture , or equivalently the switch to the moving frame , which allowed us to simplify the problem considerably , has implications beyond the quantum navigation problem considered here . in this connection
we remark that if we endow the quantum state space with the metric of the form ( [ eq:1 ] ) , then the resulting space is known as a randers space ( cf .
@xcite ) , which is an example of a finsler space ( for @xmath97 the space reduces to a kropina space ) . in the literature of finsler geometry
it is known that every zermelo navigation problem on a riemannian manifold can be solved by finding geodesics of the corresponding randers space @xcite .
the classification of spaces in terms of their curvatures is of particular importance in geometry , and in the case of randers spaces the identification of constant curvature metrics has been obtained by solving the associated navigation problem when the wind is a conformal vector field with constant scale factor @xcite ( the field generated by a unitary motion belongs to this class , with vanishing scale factor ) . however , and with hindsight , the main theorem of @xcite can be proven in essentially two lines by adopting the interaction picture , and this shows that in the present context a physical intuition can offer insights into a purely mathematical question .
we conclude by remarking that an interesting generalisation of the present problem is to consider the case under which the external influences contain noise .
in such a situation , the idea of reaching a target pure state is no longer tenable , whereas reaching a particular mixed state might be feasible .
although there appears to be a surprisingly limited amount of work in the stochastic extensions of the zermelo navigation problem ( cf .
@xcite ) , extensions into the mixed - state domain are desirable for the designing and understanding of more robust controlled quantum dynamics .
we thank eva - maria graefe , benjamin russell , and dmitry savin for discussion and comments , and martin meier for support in producing figure [ fig:2 ] .
margolus , n. & levitin , l. b. 1998 the maximum speed of dynamical evolution .
_ physica _ d**120 * * , 188 .
lloyd , s. 1999 ultimate physical limits to computation .
_ nature _ * 406 * , 1047 .
khaneja , n. , glaser , s. j. & brockett , r. 2002 sub - riemannian geometry and time optimal control of three spin systems : quantum gates and coherence transfer .
_ phys .
rev .
_ a**65 * * , 032301 .
palao , j. p. & kosloff , r. 2002 quantum computing by an optimal control algorithm for unitary transformations .
_ phys .
rev .
lett . _ * 89 * , 188301
. brody , d. c. 2003 elementary derivation for passage time .
_ j. phys .
_ a**36 * * , 5587 .
giovannetti , v. , lloyd , s. & maccone , l. 2003 quantum limits to dynamical evolution .
_ phys .
rev .
_ a**67 * * , 052109 .
schulte - herbrggen , t. , sprl , a. , khaneja , n. & glaser , s. j. 2005 optimal control - based efficient synthesis of building blocks of quantum algorithms : a perspective from network complexity towards time complexity .
_ phys .
rev .
_ a**72 * * , 042331 .
carlini , a. , hosoya , a. , koike , t. & okudaira , y. 2006 time - optimal quantum evolution .
_ phys .
rev .
lett . _ * 96 * , 060503 .
boscain , u. & mason , p. 2006
time minimal trajectories for a spin 1/2 particle in a magnetic field . _ j. math .
phys .
_ * 47 * , 062101 .
brody , d. c. & hook , d. w. 2006 on optimum hamiltonians for state transformations . _ j. phys .
_ a**39 * * , l167 .
carlini , a. , hosoya , a. , koike , t. & okudaira , y. 2007 time - optimal unitary operations .
_ phys .
rev .
_ a**75 * * , 042308 .
rezakhani , a. t. , kuo , w. j. , hamma , a. , lidar , d. a. & zanardi , p. 2009 quantum adiabatic brachistochrone .
_ phys .
rev .
lett . _ * 103 * , 080502 .
caneva , t. , murphy , m. , calarco , t. , fazio , r. , montangero , s. , giovannetti , v. & santoro , g. e. 2009 optimal control at the quantum speed limit .
_ phys .
rev .
lett . _ * 103 * , 240501 .
rojo , a. g. & bloch , a. m. 2010 the rolling sphere , the quantum spin and a simple view of the landau - zener problem .
_ amer .
j. phys . _
* 78 * , 1012 .
hegerfeldt , g. c. 2013 driving at the quantum speed limit : optimal control of a two - level system .
_ phys .
rev .
lett . _ * 111 * , 260501 .
garon , a. , glaser , s. j. & sugny , d. 2013 time - optimal control of su(2 ) quantum operations .
_ phys .
rev .
_ a**88 * * , 043422 .
russell , b. & stepney , s. 2014 applications of finsler geometry to speed limits to quantum information processing .
_ int .
j. found .
comp .
sci . _ * 25 * , 489 .
wang , x. , allegra , a. , jacobs , k. , lloyd , s. , lupo , c. & mohseni , m. 2014 quantum brachistochrone curves as geodesics : obtaining accurate control protocols for time - optimal quantum gates .
arxiv:1408.2465 zermelo , e. 1931 ber das navigationsproblem bei ruhender oder vernderlicher windverteilung .
_ ztschr .
f. angew .
math .
und mech . _
* 11 * , 114 .
carathodory , c. 1935 _ variationsrechnung und partielle differentialgleichungen erster ordnung_. ( berlin : b. g. teubner ) .
russell , b. & stepney , s. 2014 zermelo navigation and a speed limit to quantum information processing .
_ phys .
rev .
_ a**90 * * , 012303 .
russell , b. & stepney , s. 2014 zermelo navigation in the quantum brachistochrone .
arxiv:1409.2055 brody , d. c. & meier , d. m. 2014 solution to the quantum zermelo navigation problem .
_ phys .
rev .
lett . _
* 114 * ( to appear ; arxiv:1409.3204 ) .
brody , d. c. & meier , d. m. 2015 elementary solution to the time - independent quantum navigation problem .
_ j. phys .
_ a**48 * * , 055302 ( 2015 ) .
sunakawa , s. 1991 _ quantum mechanics_. ( tokyo : iwanami ) .
bao , d. , robles , c. & shen , z. 2004 zermelo navigation on riemannian manifolds .
_ j. diff .
geom . _ * 66 * , 391 .
robles , c. 2007 geodesics in randers spaces of constant curvature . _
trans . am .
math .
soc . _ * 359 * , 1633 .
anandan , j. & aharonov , y. 1990 geometry of quantum evolution .
_ phys .
rev .
lett . _ * 65 * , 1697 .
kibble , t. w. b. 1979 geometrisation of quantum mechanics .
_ commun .
math .
phys . _ * 65 * , 189 .
bouchiat , c. & gibbons , g. w. 1988 non - integrable quantum phase in the evolution of a spin-1 system : a physical consequence of the non - trivial topology of the quantum state - space .
_ j. phys .
_ france * 49 * , 187 .
brody , d. c. & hughston , l. p. 2001 geometric quantum mechanics .
_ j. geom .
phys . _ * 38 * , 19 .
uzdin , r. , gnther , u. , rahav , s. & moiseyev , n. 2012 time - dependent hamiltonians with 100% evolution speed efficiency .
_ j. phys .
_ a**45 * * , 415304 .
kobayashi , s. & nomizu , k. 1969 _ foundations of differential geometry _ , vol .
ii ( new york : john wiley & sons ) .
gibbons , g. w. & warnick , c. m. 2011 the geometry of sound rays in a wind .
_ contemp .
phys . _ * 52 * , 197 .
esfahani , p. m. , chatterjee , d. & lygeros , j. 2011 on a problem of stochastic reach - avoid set characterisation . in _
ieee conference on decision and control and european control conference_. p. 7069 . |
in the analysis of practically all available data on two - pion transitions of the @xmath3 mesons from the argus , cleo , cusb , crystal ball , belle , and _ babar _ collaborations @xmath4 ( @xmath5 , @xmath6 @xmath7 ) the contribution of multi - channel @xmath0 scattering in the final - state interactions is considered .
the analysis , which is aimed at studying the scalar mesons , is performed jointly considering the above bottomonia decays , the isoscalar s - wave processes @xmath8 , which are described in our model - independent approach based on analyticity and unitarity and using an uniformization procedure , and the charmonium decay processes @xmath9 , @xmath10 with data from the crystal ball , dm2 , mark ii , mark iii , and bes ii collaborations .
we show that the experimentally observed interesting ( even mysterious ) behavior of the @xmath0 spectra of the @xmath3-family decays , beginning from the second radial excitation and higher , a bell - shaped form in the near-@xmath0-threshold region , smooth dips about 0.6 gev in the @xmath11 , about 0.45 gev in the @xmath12 , and about 0.7 gev in the @xmath13 , and also sharp dips about 1 gev in the @xmath11 is explained by the interference between the @xmath0 scattering , @xmath14 and @xmath15 contributions to the final states of these decays ( by the constructive one in the near-@xmath0-threshold region and by the destructive one in the dip regions ) .
considering multi - channel @xmath0 scattering , we shall deal with the 3-channel case ( @xmath8 ) because it was shown @xcite that this is a minimal number of coupled channels needed for obtaining correct values of @xmath17-resonance parameters . when performing our combined analysis data for the multi - channel @xmath0 scattering
were taken from many papers ( see refs . in our paper @xcite ) .
for the decay @xmath18 data were taken from mark iii , dm2 and bes ii collaborations ; for @xmath19 from mark ii and crystal ball(80 ) ( see refs . also in @xcite ) . for
@xmath20 data were used from argus @xcite , cleo @xcite , cusb @xcite , and crystal ball @xcite collaborations ; for @xmath13 and @xmath21 from cleo @xcite ; for @xmath22 from _ babar _ @xcite and belle @xcite ; for @xmath23 from belle collaboration @xcite . the used formalism for calculating the di - meson mass distributions in the quarkonia decays is analogous to the one proposed in ref .
@xcite for the decays @xmath24 and @xmath25 ( @xmath26 ) but with allowing for also amplitudes of transitions between the @xmath0 , @xmath1 and @xmath2 channels in decay formulas .
there was assumed that the pion pairs in the final state have zero isospin and spin .
only these pairs of pions undergo final state interactions whereas the final @xmath27 meson ( @xmath28 ) remains a spectator .
the amplitudes of decays are related with the scattering amplitudes @xmath29 @xmath30 as follows @xmath31 where @xmath32 , @xmath33 and @xmath34 ; indices @xmath35 and @xmath36 correspond to @xmath37 and @xmath27 , respectively .
the free parameters @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath42 , @xmath43 , @xmath44 and @xmath45 depend on the couplings of @xmath46 , @xmath47 and the @xmath37 to the channels @xmath0 , @xmath1 and @xmath2 .
the pole term in eq.(1 ) in front of @xmath48 is an approximation of possible @xmath49 states , not forbidden by ozi rules .
the amplitudes @xmath29 are expressed through the @xmath50-matrix elements @xmath51 where @xmath52 and @xmath53 is the reaction threshold .
the @xmath50-matrix elements are parameterized on the uniformization plane of the @xmath0 scattering amplitude by poles and zeros which represent resonances .
the uniformization plane is obtained by a conformal map of the 8-sheeted riemann surface , on which the three - channel @xmath50 matrix is determined , onto the plane . in the uniformizing variable
used we have neglected the @xmath0-threshold branch point and allowed for the @xmath1- and @xmath2-threshold branch points and left - hand branch point at @xmath54 related to the crossed channels .
the background is introduced to the amplitudes in a natural way : on the threshold of each important channel there appears generally speaking a complex phase shift .
it is important that we have obtained practically zero background of the @xmath0 scattering in the scalar - isoscalar channel .
it confirms well our representation of resonances .
the expression @xmath55[m_\psi^2- ( \sqrt{s}+m_\phi)^2]}$ ] for decay @xmath9 and the analogues relations for @xmath56 and @xmath4 give the di - meson mass distributions .
@xmath57 ( normalization to experiment ) is : for @xmath9 0.5172 ( mark iii ) , 0.1746 ( dm 2 ) and 3.8 ( bes ii ) ; for @xmath58 1.746 ( mark
ii ) ; for @xmath59 1.6891 ( crystal ball(80 ) ) ; for @xmath60 4.1758 ( argus ) , 2.0445 ( cleo(94 ) ) and 1.0782 ( cusb ) ; for @xmath61 0.0761 ( crystal ball(85 ) ) ; for @xmath62 19.8825 and 4.622 ( cleo(07 ) ) ; for @xmath63 @xmath64 1.6987 and 1.1803 ( cleo(94 ) ) ; for @xmath65 4.6827 ( _ babar_(06 ) ) and 0.3636 ( belle(07 ) ) ; for @xmath66 , 37.9877 ( _ babar_(06 ) ) ; for @xmath67 , @xmath68 and @xmath69 respectively 0.2047 , 2.8376 and 6.9251 ( belle(12 ) ) . satisfactory combined description of all considered processes ( including @xmath8 ) is obtained with the total @xmath70 ; for the @xmath0 scattering , @xmath71 . studying the decays of charmonia and bottomonia , we investigated the role of the individual @xmath17 resonances in contributing to the shape of the dipion mass distributions . in this case
we switched off only those resonances [ @xmath72 , @xmath73 , @xmath74 and @xmath75 , removal of which can be somehow compensated by correcting the background to have the more - or - less acceptable description of the multichannel @xmath0 scattering .
first , when leaving out before - mentioned resonances , a minimal set of the @xmath17 mesons consisting of the @xmath72 , @xmath76 , and @xmath77 is sufficient to achieve a description of the processes @xmath78 with a total @xmath79 .
second , from these three mesons only the @xmath72 can be switched off while still obtaining a reasonable description of multi - channel @xmath0-scattering ( though with an appearance of the pseudo - background ) with a total @xmath80 . in figures 1 - 3
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 . and
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=50.0% ] + and @xmath10 .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ] and @xmath10 .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ] [ fig : besii ] ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 .
, title="fig:",scaledwidth=45.0% ] ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 .
, title="fig:",scaledwidth=45.0% ] + ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 . , title="fig:",scaledwidth=45.0% ] ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 .
, title="fig:",scaledwidth=45.0% ] + ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 .
, title="fig:",scaledwidth=45.0% ] ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 .
, title="fig:",scaledwidth=45.0% ] + ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 .
, title="fig:",scaledwidth=45.0% ] ( two upper panels ) , @xmath81 ( middle panel ) and @xmath82 ( lower panel ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77 . ,
title="fig:",scaledwidth=45.0% ] [ fig : ups21 ] ( left - hand ) and @xmath83 ( @xmath7 ) ( right - hand ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ] ( left - hand ) and @xmath83 ( @xmath7 ) ( right - hand ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ] + ( left - hand ) and @xmath83 ( @xmath7 ) ( right - hand ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ] ( left - hand ) and @xmath83 ( @xmath7 ) ( right - hand ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ] + ( left - hand ) and @xmath83 ( @xmath7 ) ( right - hand ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ] ( left - hand ) and @xmath83 ( @xmath7 ) ( right - hand ) .
the solid lines correspond to contribution of all relevant @xmath17-resonances ; the dotted , of the @xmath72 , @xmath76 , and @xmath77 ; the dashed , of the @xmath76 and @xmath77.,title="fig:",scaledwidth=45.0% ]
the combined analysis was performed for data on isoscalar s - wave processes @xmath8 and on the decays of the charmonia @xmath9 , @xmath84 and of the bottomonia @xmath4 ( @xmath5 , @xmath6 @xmath7 ) from the argus , crystal ball , cleo , cusb , dm2 , mark ii , mark iii , bes ii , _ babar _ , and belle collaborations .
it is shown that the dipion mass spectra in the above - indicated decays of charmonia and bottomonia are explained by the unified mechanism which is based on our previous conclusions on wide resonances @xcite and is related to contributions of the @xmath0 , @xmath1 and @xmath2 coupled channels including their interference .
it is shown that in the final states of these decays ( except @xmath0 scattering ) the contribution of coupled processes , e.g. , @xmath85 , is important even if these processes are energetically forbidden .
the role of the individual @xmath17 resonances in making up the shape of the dipion mass distributions in the charmonia and bottomonia decays is considered .
note the unexpected result a considerable contribution of the @xmath73 to the bell - shaped form of the dipion mass spectra of bottomonia decays in the near-@xmath0-threshold region .
since describing the bottomonia decays , we did not change resonance parameters in comparison with the ones obtained in the combined analysis of the processes @xmath8 and charmonia decays , the results of this analysis confirm all of our earlier conclusions on the scalar mesons @xcite .
+ this work was supported in part by the heisenberg - landau program , by the votruba - blokhintsev program for cooperation of czech republic with jinr , by the grant agency of the czech republic ( grant no .
p203/15/04301 ) , by the grant program of plenipotentiary of slovak republic at jinr , by the bogoliubov - infeld program for cooperation of poland with jinr , by the bmbf ( project 05p2015 , bmbf - fsp 202 ) , by tomsk state university competitiveness improvement program , the russian federation program `` nauka '' ( contract no . 0.1526.2015 , 3854 ) , by slovak grant agency vega under contract no.2/0197/14 , and by the polish national science center ( ncn ) grant dec-2013/09/b / st2/04382 . | it is shown that the basic shape of dipion mass distributions in the two - pion transitions of both charmonia and bottomonia states are explained by an unified mechanism based on the contribution of the @xmath0 , @xmath1 and @xmath2 coupled channels including their interference . |
in this paper we give some generalization of the approach suggested in @xcite , which is aimed to reduce equations with variable coefficients to more simple ones .
this transform was used in a series of papers @xcite , @xcite-@xcite to investigate in a unified way several equations such as the linear and semilinear tricomi equations , gellerstedt equation , the wave equation in einstein - de sitter spacetime , the wave and the klein - gordon equations in the de sitter and anti - de sitter spacetimes .
the listed equations play an important role in the gas dynamics , elementary particle physics , quantum field theory in curved spaces , and cosmology .
consider for the smooth function @xmath0 the solution @xmath1 to the problem @xmath2\subseteq { \mathbb r } , \,\ , x \in \omega \subseteq { \mathbb r}^n,\ ] ] with the parameter @xmath3 \subseteq { \mathbb r}$ ] , @xmath4 , and with @xmath5 . here
@xmath6 is a domain in @xmath7 , while @xmath8 is the partial differential operator @xmath9 .
we are going to present the integral operator @xmath10(x , t ) = 2 \int _ { t_0}^{t } db \int _ { 0}^ { |\phi ( t)- \phi ( b)| } k ( t;r , b;m ) w(x , r;b ) dr , \quad x \in \omega , \,\ , t \in i,\ ] ] which maps the function @xmath11 into solution of the equation @xmath12 in fact , the function @xmath13 takes initial values as follows @xmath14 here @xmath15 is a distance function produced by @xmath16 , that is @xmath17 , while @xmath18 is a constant .
moreover , we also give the corresponding operators , which generate solutions of the source - free equation and takes non - vanishing initial values .
these operators are constructed in @xcite in the case of @xmath19 , where @xmath20 is the laplace operator on @xmath7 , and , consequently , the equation ( [ 0.1jde ] ) is the wave equation . in the present paper
we restrict ourselves to the smooth functions , but it is evident that similar formulas , with the corresponding interpretations , are applicable to the distributions as well .
( for details see , e.g. , @xcite . ) in order to motivate our approach , we consider the solution @xmath21 to the cauchy problem @xmath22 with the parameter @xmath23 .
we denote that solution by @xmath24 ; if @xmath25 is independent of the second time variable @xmath26 , then we write simply @xmath27 .
there are well - known explicit representation formulas for the solution of the problem ( [ 1.7new ] ) .
( see , e.g. , @xcite . )
the starting point of the approach suggested in @xcite is the duhamel s principle ( see , e.g. , @xcite ) , which has been revised in order to prepare the ground for generalization .
our _ first observation _ is that the function @xmath28 is the solution of the cauchy problem @xmath29 in @xmath30 , and @xmath31 , @xmath32 in @xmath33 if the function @xmath34 is a solution of the problem ( [ 1.7new ] ) , where @xmath35 .
the _ second observation _ is that in ( [ main ] ) the upper limit @xmath36 of the inner integral is generated by the propagation phenomena with the speed which equals to one .
in fact , that is a distance function .
our _ third observation _ is that the solution operator @xmath37 can be regarded as a composition of two operators .
the first one @xmath38 is a fourier integral operator , which is a solution operator of the cauchy problem for wave equation .
the second operator @xmath39 is the integral operator given by ( [ main ] ) .
we regard the variable @xmath26 in ( [ main ] ) as a `` subsidiary time '' .
thus , @xmath40 and we arrive at the diagram of figure 1 . case of general @xmath41,title="fig:",scaledwidth=25.0% ] case of general @xmath41,title="fig:",scaledwidth=25.0% ] based on the first diagram , we have generated in @xcite a class of operators for which we have obtained explicit representation formulas for the solutions , and , in particular , the representations for the _ fundamental solutions of the partial differential operator_. in fact , this diagram brings into a single hierarchy several different partial differential operators . indeed , if we take into account the propagation cone by introducing the distance function @xmath42 , and if we provide the integral operator ( [ main ] ) with the kernel @xmath43 , as in ( [ kjde ] ) , then we actually generate new representations for the solutions of different well - known equations with @xmath44-independent coefficients .
( see , for details , @xcite . ) in the present paper we extend the class of the equations for which we can obtain explicit representation formulas for the solutions , by varying the first mapping .
more precisely , consider the diagram of figure 2 , where @xmath45 is a solution to the problem ( [ 0.1jde ] ) with the parameter @xmath23 .
if we have a resolving operator of the problem ( [ 0.1jde ] ) , then , by applying ( [ kjde ] ) , we can generate solutions of another equation .
thus , @xmath46 .
the new class of equations contains operators with @xmath44-depending coefficients , and those equations are not necessarily hyperbolic . in this paper
we restrict ourselves to the klein - gordon equation in the de sitter spacetime , that is @xmath47 in ( [ 0.1jde ] ) .
recently the equations in the de sitter and anti - de sitter spacetimes became the focus of interest for an increasing number of authors ( see , e.g. , @xcite-@xcite , @xcite-@xcite , @xcite , @xcite-@xcite , @xcite -@xcite , @xcite and the bibliography therein ) which investigate those equations from a wide spectrum of perspectives .
the creation of a tool for the investigation of the local and global solvability in the problems for these linear and nonlinear equations appears to be a worthwhile undertaking .
we believe that the integral transform and the representation formulas for the solutions that we derive in this article fill up the gap in the literature on that topic .
to formulate the main result of this paper we need the following notations .
first , we define a _ chronological future _ @xmath48 and a _ chronological past _ @xmath49 of the point @xmath50 , @xmath51 , @xmath52 , as follows : @xmath53 .
then , for @xmath54 , @xmath55 , we define the function @xmath56 where @xmath57 and @xmath58 is the hypergeometric function .
( for definition of the hypergeometric function , see , e.g. , @xcite . ) when no ambiguity arises , like in ( [ em ] ) , we use the notation @xmath59 for @xmath60 .
thus , the function @xmath61 depends on @xmath62 , that is @xmath63 . according to theorem [ t2.12 ]
, the function @xmath64 solves the following one dimensional klein - gordon equation in the de sitter spacetime : @xmath65 the kernels @xmath66 and @xmath66 are defined by @xmath67\sqrt{(1+e^{-t } ) ^2 - z^2 } } \\ & & \times \bigg [ \big ( e^{-t } -1 + m(e^ { -2 t } - 1 - z^2 ) \big ) f \big(\frac{1}{2}-m , \frac{1}{2}-m ; 1 ; \frac { ( 1-e^{-t } ) ^2 -z^2 } { ( 1+e^{-t } ) ^2 -z^2 } \big)\nonumber \\ & & \hspace{0.3 cm } + \big ( 1-e^{-2 t}+ z^2 \big)\big ( \frac{1}{2}+m\big ) f \big(-\frac{1}{2}-m , \frac{1}{2}-m ; 1 ; \frac { ( 1-e^{-t } ) ^2 -z^2 } { ( 1+e^{-t } ) ^2 -z^2 } \big ) \bigg]\nonumber , \\
\label{k1 m } k_1(z , t;m ) & : = & 4 ^{-m } e^ { mt } \big((1+e^{-t } ) ^2 - z ^2\big)^{-\frac{1}{2}+m } f\left(\frac{1}{2}-m , \frac{1}{2}-m ; 1 ; \frac { ( 1-e^{-t } ) ^2 -z^2 } { ( 1+e^{-t } ) ^2 -z^2 } \right)\ , .
\end{aligned}\ ] ] the equation ( [ 0.3jde ] ) is said to be an equation with imaginary ( real ) mass if there is minus ( plus ) in front of @xmath68 ; however @xmath55 .
for the klein - gordon equation with imaginary mass we have the following result . [ t1.2 ] for @xmath69 , @xmath70 $ ] , @xmath71 , and @xmath72 , @xmath73 ,
let the function @xmath74 be a solution to the problem @xmath75\,,\cr v(x,0;b)=f(x , b)\ , , \quad v_t(x,0;b)= 0\ , , \quad b \in i,\quad x \in \omega\,,}\ ] ] and the function @xmath76 be a solution of the problem @xmath77\ , , \cr v(x,0)= \varphi ( x ) , \quad v_t(x,0)=0\,,\quad x \in \omega\,.}\ ] ] then the function @xmath78 defined by @xmath79 where @xmath80 , solves the problem @xmath81 here the kernels @xmath82 , @xmath83 and @xmath84 have been defined in ( [ em ] ) , ( [ k0 m ] ) and ( [ k1 m ] ) , respectively .
we note that the operator @xmath41 is of arbitrary order , that is , the equation of ( [ 1.20 ] ) can be an evolution equation , not necessarily hyperbolic .
then , the problems in ( [ 1.22 ] ) and ( [ 1.20 ] ) can be a mixed initial - boundary value problem involving the boundary condition .
indeed , assume that @xmath85 is domain with smooth boundary @xmath86 , and that @xmath87 is a unit normal vector .
let @xmath88 and @xmath89 be continuous functions , @xmath90 . if @xmath91 satisfies the boundary condition @xmath92 , \quad x \in \partial \omega \,,\ ] ] then the function @xmath78 fulfills the same boundary condition @xmath93 next , we stress that interval @xmath94\subseteq [ 0,1]$ ] , which appears in ( [ 1.22 ] ) , reflects the fact that de sitter model possesses the horizon @xcite ; existence of the horizon in the de sitter model is widely used to define an asymptotically de sitter space ( see , e.g. , @xcite ) and to involve geometry into the analysis of the operators on the de sitter space ( see , e.g. , @xcite ) .
although the next statement is a straightforward consequence of theorem [ t1.2 ] , we present it as a separate theorem having in mind the importance of the equation with real mass . in the next theorem
we use the following kernels @xmath95 which were introduced in @xcite .
[ t1.1 ] for @xmath69 , @xmath70 $ ] , @xmath71 , and @xmath72 , @xmath73 , let the function @xmath74 be a solution to the problem @xmath96\,,\cr v(x,0;b)=f(x , b)\ , , \quad v_t(x,0;b)= 0\ , , \quad b \in i,\quad x \in \omega\,,}\ ] ] and the function @xmath97 be a solution of the problem @xmath98\,,\cr v(x,0)= \varphi ( x ) , \quad v_t(x,0)=0\,,\quad x \in \omega\,.}\ ] ] then the function @xmath78 defined by @xmath99 where @xmath80 , solves the problem @xmath100 here the kernels @xmath82 , @xmath83 and @xmath84 have been defined in ( [ ea ] ) , ( [ k0 ] ) and ( [ k1 ] ) , respectively . the special cases of theorems [ t1.2 ] , [ t1.1 ] , when @xmath19 , one can find in @xcite .
the proof given in those papers is based on the well - known explicit representation formulas for the wave equation , the riemann function , the spherical means , and the asgeirsson s mean value theorem .
the main outcome , resulting from the application of all those tools , is the derivation of the final representation formula and the kernels @xmath82 , @xmath83 , and @xmath84 .
having in the hand the integral transform and the final formulas , we suggest here straightforward proof by substitution , which works also for the equations with coefficients depending on @xmath44
. among possible applications of the integral transform method are the @xmath101 estimates , strichartz estimates , huygens principle , global and local existence theorem for semilinear and quasilinear equations .
below we give examples of the equations with the variable coefficients those are amenable to the integral transform method .
* example 1 . * the metric @xmath102 in the de sitter type spacetime , that is , @xmath103 , @xmath104 , @xmath105 , @xmath106 , @xmath107 , @xmath108 , where @xmath109 , and @xmath110 is kronecker s delta .
the linear covariant klein - gordon equation in the coordinates is @xmath111 here @xmath112 is a physical mass of the particle .
if we introduce the new unknown function @xmath113 , then the equation takes the form of the klein - gordon equation with imaginary mass @xmath114 where @xmath115 is the square of the so - called curved ( or effective ) mass . for the last equation we set @xmath116
if @xmath117 is a non - euclidean space of constant negative curvature and the equation of the problems ( [ 1.22 ] ) and ( [ 1.23 ] ) is a non - euclidean wave equation , then the explicit representation formulas are known ( see , e.g. , @xcite ) and the huygens principle is a consequence of those formulas .
thus , for a non - euclidean wave equation , due to theorem [ t1.2 ] , the functions @xmath118 and @xmath119 have explicit representations , and the arguments of @xcite allow us to derive for the solution @xmath120 of the problem ( [ 1.20 ] ) in the de sitter type metric with hyperbolic spatial geometry the explicit representation , the @xmath101 estimates , and to examine the huygens principle .
precise statements will be published in the forthcoming paper .
* example 2 . * the euler - bernoulli beam equation with the variable coefficients @xmath121 here @xmath122 and
we assume that @xmath123 .
this paper is organized as follows . in section [ s2 ]
we study the kernel functions @xmath82 , @xmath83 , and @xmath84 , and prove several basic properties of those function . then , in section [ s3 ] , we prove theorem [ t1.2 ] . applications of theorems [ t1.2],[t1.1 ] to some equations appearing in electrodynamics and cosmology will be done in a forthcoming paper .
the proof of theorem [ t1.2 ] is straightforward ; we just substitute the function of ( [ 1.21 ] ) in the equation of ( [ 1.20 ] ) , and then check the initial conditions .
it is straightforward , but not short ; it requires very long and tedious calculations . in order to make the calculations more transparent ,
we reveal in this section several main properties of the kernel functions . in this subsection
we collect some important properties of the kernel @xmath125 .
although the function @xmath125 ( [ em ] ) is defined for @xmath126 , we use it for @xmath127 and @xmath128 , only . consider for @xmath129 , @xmath130 , @xmath55 , @xmath131 $ ] the function @xmath132 the following notations are helpful to evaluate derivatives of the kernel functions @xmath133 we rewrite the function @xmath134 in terms of these functions as follows @xmath135 for the derivatives of the auxiliary functions @xmath136 , @xmath137 , and @xmath138 we have [ l2.2new ] the partial derivatives of the functions @xmath136 , @xmath137 , and @xmath138 are as follows @xmath139\,,\\ \alpha_{rr } ( r , t , b;m ) & = & -2 m \alpha ( r , t , b;m ) \beta^2 ( r , t , b)\left[1 - 2 m r^2 \beta^2 ( r , t , b)+4 r \beta^2 ( r , t , b)\right ] \,,\\
\beta_{rr } ( r , t , b ) & = & \beta^3 ( r , t , b)+ 3r^2 \beta^5 ( r , t , b)\,,\\ \gamma_{rr } ( r , t , b ) & = & 2 \beta^2 ( r , t , b)(1 + 4r^2\beta^2 ( r , t , b ) ) \left [ \gamma ( r , t , b ) - 1\right]\ , .
\end{aligned}\ ] ] * proof . * we skip the simple proof of formulas for the first - order derivatives and of @xmath140 . for the second - order derivatives we have @xmath141\ , , \\
\gamma_{rr } ( r , t , b ) & = & 2 \beta^2 ( r , t , b)\left [ \gamma ( r , t , b ) - 1\right ] + 4 r \beta_r ( r , t , b)\beta ( r , t , b)\left [ \gamma ( r , t , b ) - 1\right ] \\ & & + 2 r \beta^2 ( r , t , b ) \gamma_r ( r , t , b ) \\ & = & 2 \beta^2 ( r , t , b)\left [ \gamma ( r , t , b ) - 1\right ] + 4 r^2 \beta^4 ( r , t , b)\left [ \gamma ( r , t , b ) - 1\right ] \\ & & + 4 r^2 \beta^4 ( r , t , b)\left [ \gamma ( r , t , b ) - 1\right ] \\ & = & 2 \beta^2 ( r , t , b)(1 + 4r^2\beta^2 ( r , t , b ) ) \left [ \gamma ( r , t , b ) - 1\right ] .
\end{aligned}\ ] ] lemma is proven .
@xmath142 [ c1.2 ]
the derivatives , explicitly written , are as follows : @xmath143 in particular , for @xmath144 we obtain the next values of those functions .
[ l2.2 ] for the functions @xmath136 , @xmath145 , and @xmath138 we have @xmath146 while for their derivatives we have @xmath147 * proof . *
it is a simple consequence of the definitions of these functions and the previous lemma .
lemma is proven .
@xmath142 now we turn to the derivatives of the function @xmath148 .
[ p2.3 ] the derivative @xmath149 is given as follows @xmath150\,.\end{aligned}\ ] ] moreover , @xmath151 the second - order derivative @xmath152 is given by @xmath153 \\ & & \times \bigg [ f \big(\frac{1}{2}-m , \frac{1}{2}-m ; 1;\gamma ( r , t , b)\big ) \\ & & \hspace{0.5 cm } + ( \gamma ( r , t , b ) - 1 ) \left(\frac{1}{2}-m\right ) f\left(\frac{3}{2}-m,\frac{3}{2}-m;2;\gamma ( r , t , b)\right ) \bigg ] \\ & & + \,4\left(\frac{1}{2}-m\right)^2 \left(\frac{3}{2}-m\right)r^2 \alpha ( r , t , b;m ) \beta^5 ( r , t , b)\left ( \gamma ( r , t , b ) - 1\right ) \\ & & \times \bigg [ f \big(\frac{3}{2}-m , \frac{3}{2}-m ; 2;\gamma ( r , t , b)\big ) \\ & & \hspace{0.5 cm }
+ ( \gamma ( r , t , b ) - 1 ) \left(\frac{3}{2}-m\right ) \frac{1}{2 } f \left(\frac{5}{2}-m,\frac{5}{2}-m;3;\gamma ( r , t , b)\right)\bigg]\,.\end{aligned}\ ] ] * proof .
* indeed , due to ( * ? ? ?
* ( 7 ) sec .
2.1.2 ) , we have @xmath154f \big(\frac{1}{2}-m , \frac{1}{2}-m ; 1;\gamma ( r , t , b)\big)\\ & & + \alpha ( r , t , b;m ) \beta ( r , t , b ) \gamma_r ( r , t , b)\left(\frac{1}{2}-m\right)^2 f\left(\frac{3}{2}-m,\frac{3}{2}-m;2;\gamma ( r , t , b)\right)\\ & = & \bigg [ -2 mr \alpha ( r , t , b;m ) \beta^3 ( r , t , b ) \\ & & + \alpha ( r , t , b;m ) r \beta^3 ( r , t , b ) \bigg]f \big(\frac{1}{2}-m , \frac{1}{2}-m ; 1;\gamma ( r , t , b)\big)\\ & & + 2 r \alpha ( r , t , b;m ) \beta^3 ( r , t , b)\left [ \gamma ( r , t , b ) - 1\right ] \left(\frac{1}{2}-m\right)^2 f\left(\frac{3}{2}-m,\frac{3}{2}-m;2;\gamma ( r , t , b)\right ) \,.\end{aligned}\ ] ] this proves the first formula of the lemma and the statement for @xmath155 .
then , for the second - order derivative @xmath156 we have @xmath157 \\ & & + 4\left(\frac{1}{2}-m\right)r^2 \alpha ( r , t , b;m ) \beta^5 ( r , t , b)\left [ \gamma ( r , t , b ) - 1\right]\\ & & \times \bigg [ f_z \big(\frac{1}{2}-m , \frac{1}{2}-m ; 1;\gamma ( r , t , b)\big ) + \left(\frac{1}{2}-m\right ) f\left(\frac{3}{2}-m,\frac{3}{2}-m;2;\gamma ( r , t , b)\right ) \\ &
& + ( \gamma ( r , t , b ) - 1 ) \left(\frac{1}{2}-m\right ) f_z \left(\frac{3}{2}-m,\frac{3}{2}-m;2;\gamma ( r , t , b)\right)\bigg ] \\ & = & 2\left(\frac{1}{2}-m\right ) \big \ { \alpha ( r , t , b;m ) \beta^3 ( r , t , b ) -2 mr^2 \alpha ( r , t , b;m ) \beta^5 ( r , t , b ) + 3r^2 \alpha ( r , t , b;m ) \beta^5 ( r , t , b)\big \ } \\ & & \times \bigg [ f \big(\frac{1}{2}-m , \frac{1}{2}-m ; 1;\gamma ( r , t , b)\big ) \\ & & + ( \gamma ( r , t , b ) - 1 ) \left(\frac{1}{2}-m\right ) f\left(\frac{3}{2}-m,\frac{3}{2}-m;2;\gamma ( r , t , b)\right ) \bigg ] \\ & & + 4\left(\frac{1}{2}-m\right)^2 \left(\frac{3}{2}-m\right)r^2 \alpha ( r , t , b;m ) \beta^5 ( r , t , b)\left ( \gamma ( r , t , b ) - 1\right ) \\ & & \times \bigg [ f \big(\frac{3}{2}-m , \frac{3}{2}-m ; 2;\gamma ( r , t , b)\big ) \\ & & + ( \gamma ( r , t , b ) - 1 ) \left(\frac{3}{2}-m\right ) \frac{1}{2 } f \left(\frac{5}{2}-m,\frac{5}{2}-m;3;\gamma ( r , t , b)\right)\bigg ] \,.\end{aligned}\ ] ] it is easily seen that the last expression coincides with one given in the statement of the proposition .
this completes the proof of the proposition .
@xmath142 in fact , corollary [ c1.2 ] allows us to write @xmath158 and @xmath156 in the explicit form .
[ c2.4 ] we have @xmath159\,,\\ \partial_r^2 e(r , t ; 0,b;m ) & = & \widetilde{a}(r , t , b;m ) f \big(\frac{1}{2}-m,\frac{1}{2}-m;1;\gamma ( r , t , b ) \big)\\ & & + \widetilde{b}(r , t , b;m ) f \big(\frac{3}{2}-m,\frac{3}{2}-m;2;\gamma ( r , t , b ) \big)\\ & & + \widetilde{c}(r , t , b;m ) f \big(\frac{5}{2}-m,\frac{5}{2}-m;3;\gamma ( r , t , b ) \big)\,,\end{aligned}\ ] ] where @xmath160 \,,\\ \widetilde{b}(r , t , b;m ) & = & -2^{3 - 2 m}\left(\frac{1}{2}-m\right)^2 e^{(m-1 ) ( b+t)}\left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{7}{2 } } \left [ \left(e^{-b}+e^{-t}\right)^2 + \left(5 - 4m\right ) r^2 \right ] , \\ \widetilde{c}(r , t , b;m ) & = & 2^{5 - 2 m } \left(\frac{1}{2}-m\right)^2 \left(\frac{3}{2}-m\right)^2 e^{(m -2)(b+t ) } r^2 \left(\left(e^{-b}+e^{-t}\right)^2 -r^2\right)^{m-\frac{9}{2}}\,.\\\end{aligned}\ ] ] next , we turn to the derivatives with respect to time . we skip the proof of the following lemma , which gives the derivatives of the auxiliary functions @xmath161 , @xmath137 , and @xmath138 .
[ l2.6 ] for the first - order derivatives of the functions @xmath161 , @xmath137 , and @xmath138 we have @xmath162 and @xmath163 [ p2.7 ] one can write @xmath164 where @xmath165 * proof . *
indeed , it is easily seen that @xmath166 and , consequently , @xmath167 proposition is proven .
@xmath142 [ c2.8 ] the coefficients @xmath168 and @xmath169 , explicitly written , are as follows @xmath170 \,,\\ b(r , t , b;m ) & = & 2^{1 - 2 m } \left(\frac{1}{2}-m\right)^2 e^{m ( b+t ) } e^{-t } \left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{5}{2}}\\ & & \times \left [ \left(e^{-b}-e^{-t}\right ) \left(e^{-b}+e^{-t}\right)^2-\left(e^{-b}-e^{-t}\right)r^2 \right.\\ & & \left.+ \left(e^{-b}+e^{-t}\right ) \left(e^{-b}-e^{-t}\right)^2-\left(e^{-b}+e^{-t}\right ) r^2 \right ] \,.\end{aligned}\ ] ] in particular , @xmath171 then , the following proposition will be used .
[ p2.9 ] for all @xmath172 , @xmath173 , we have @xmath174 * proof . *
according to proposition [ p2.3 ] and ( [ 2.2 ] ) , we can write @xmath175 proposition is proven .
@xmath176 [ p2.10 ] for all @xmath177 , @xmath173 , we have @xmath178 where @xmath179 * proof .
* it can be verified by simple calculations .
proposition is proven .
@xmath142 [ c2.11 ]
the functions @xmath180 , @xmath181 , and @xmath182 , explicitly written , are as follows @xmath183\\ & & + 2^{2 - 2 m } \left(m-\frac{1}{2}\right ) \left(m-\frac{3}{2}\right ) \left(e^{-b}+e^{-t}\right)^2 e^{m ( b+t)-2 t } \left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{5}{2}}\,,\\ & & c ( r , t , b;m)\\ & = & -2^{3 - 2 m } \left(\frac{1}{2}-m\right)^2 m e^{m ( b+t)-3 b - t } \left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{5}{2}}\left[-e^{2 b } r^2-e^{2 b-2 t}+1\right ] \\ & & -2^{2 - 2 m } \left(\frac{1}{2}-m\right)^2 e^{(m-5 ) ( b+t)}\left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{7}{2}}\\ & & \times \big [ -4 m r^2 e^{3 b+3 t}-4 m r^2 e^{4b+2 t } + 4 m e^{2 ( b+t)}-4 m e^{3 b+t}+4 m e^{b+3
t}-4 e^{4 b } m\\ & & + r^4 e^{4 b+4 t}+4 r^2 e^{3 b+3 t}-2 r^2 e^{2 b+4 t}+8 r^2 e^{4b+2 t } -8 e^{2 ( b+t)}-4 e^{b+3 t}+3 e^{4 b}+e^{4 t } \big]\,,\\ & & d(r , t , b;m ) \\ & = & 2^{3 - 2 m } \left(\frac{1}{2}-m\right)^2 \left(\frac{3}{2}-m\right)^2 e^{m ( b+t ) -2 t } \left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{9}{2}}\\ & & \times \left[e^{-2 b } r^4 + 2 r^2 e^{-2 b-2 t}-2 e^{-4 b } r^2 - 2 e^{-4 b-2 t}+e^{-2 b-4 t}+e^{-6 b } \right ] \,.\end{aligned}\ ] ] [ t2.12 ] the function @xmath184 solves the following equation @xmath185 * proof . *
we have @xmath186 where @xmath187 that is , @xmath188\\ & & + 2^{2 - 2 m } \left(m-\frac{1}{2}\right ) \left(m-\frac{3}{2}\right ) \left(e^{-b}+e^{-t}\right)^2 e^{m ( b+t)-2 t } \left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{5}{2 } } \\ & & -e^{-2 t}\left(2^{2 - 2 m } \left(m-\frac{3}{2}\right ) \left(m-\frac{1}{2}\right ) r^2 e^{m ( b+t ) } \left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{5}{2 } } \right.\\ & & -\left.2^{1 - 2 m } \left(m-\frac{1}{2}\right ) e^{m ( b+t ) } \left(\left(e^{-b}+e^{-t}\right)^2-r^2\right)^{m-\frac{3}{2}}\right)\,.\end{aligned}\ ] ] after simplification we obtain @xmath189 then @xmath190 after simplification we obtain @xmath191\,.\end{aligned}\ ] ] then @xmath192 in fact , we obtain @xmath193\,.\end{aligned}\ ] ] furthermore , @xmath194 where @xmath195 and the following notations have been used @xmath196 hence @xmath197 next , if we denote @xmath198 then @xmath199 since @xmath200 the hypergeometric function @xmath201 solves the following equation @xmath202 theorem is proven .
@xmath142 by definition @xmath204 where @xmath205 , @xmath206 $ ] , and @xmath207 .
in fact , @xmath208 which simplifies the proof of many properties of this function since they are inherited from the kernel @xmath209 . [ p2.13 ] we have @xmath210 * proof . * due to corollary [ c2.4 ] , the derivative @xmath211 , explicitly written , is as follows @xmath212 \nonumber \,.\end{aligned}\ ] ] in particular , this proves the second equation of the proposition .
moreover , if @xmath213 , then @xmath214 then , due to proposition [ p2.7 ] and corollary [ c2.8 ] , the derivative @xmath215 is @xmath216 where @xmath217 and @xmath218 hence , @xmath219 in particular , @xmath220 while @xmath221 this completes the proof of the proposition .
@xmath142 furthermore , the following proposition easily follows from theorem [ t2.12 ] .
[ p2.14 ] the function @xmath222 solves the following equation @xmath223 the function @xmath225 can be written as follows @xmath226_{b=0}\,.\end{aligned}\ ] ] we consider @xmath225 for @xmath227 , and @xmath55 .
we have @xmath228 then @xmath229 f \left(\frac{1}{2}-m,\frac{1}{2}-m;1;\gamma ( r , t,0)\right ) \\
& & -\frac{2 r^4 e^{5 t}+12 r^2 e^{3 t}-4 r^2 e^{5 t}-14 e^t+12 e^{3 t}+2 e^{5 t } } { e^t \left(\left(r^2 - 1\right ) e^t-2\right)-1 } f \left(\frac{1}{2}-m,\frac{3}{2}-m;1;\gamma ( r , t,0)\right)\bigg\}\,.\end{aligned}\ ] ] thus , @xmath230 similarly , @xmath231 f\left(\frac{1}{2}-m,\frac{1}{2}-m;1;\frac{\left(1-e^{-t}\right)^2-r^2}{\left(e^{-t}+1\right)^2-r^2}\right)\nonumber\\ & & -\frac{1}{e^t \left(\left(r^2 - 1\right ) e^t-2\right)-1}\bigg[4 m r^4 e^{5 t}+4 m r^4 e^{7 t}-8 m r^2 e^{3 t}+48 m r^2 e^{5 t}\nonumber\\ & & -8 m r^2 e^{7 t}+4 m e^t-4 m e^{3 t}-4 m e^{5 t}+4 m
e^{7 t}-2 r^4 e^{5 t}-2 r^4 e^{7 t}+4 r^2 e^{3 t}\nonumber\\ & & -24 r^2 e^{5 t}+4 r^2 e^{7 t}-2 e^t+2 e^{3 t}+2 e^{5 t}-2 e^{7 t}\bigg ] f\left(\frac{1}{2}-m,\frac{3}{2}-m;1;\frac{\left(1-e^{-t}\right)^2-r^2}{\left(e^{-t}+1\right)^2-r^2}\right ) \bigg\}\nonumber\end{aligned}\ ] ] and @xmath232 [ p2.15 ] for all @xmath233 we have @xmath234\,,\\ & & \left ( \frac{1}{4}- m^2\right ) e ^{\frac{t}{2 } } - 2 e ^{-t } k_0(\phi ( t),t;m ) + 4 e ^{-2 t } k_{0\,r}(\phi ( t),t;m ) + 4 e ^{-t } k_{0\,t}(\phi ( t),t;m ) = 0\,.\end{aligned}\ ] ] * proof . *
the first and second statements follow from the corresponding ones for the function @xmath82 . then , according to ( [ 1.8jde ] ) , ( [ 1.9jde ] ) , and ( [ 1.10jde ] ) , we have @xmath235 \\ & & + 4 e ^{-2 t } \bigg [ -\frac{1}{16 } m^4 e^ { \frac{1}{2}t}+\frac{1}{8 } m^4 e^ { \frac{3}{2 } t}-\frac{1}{16 } m^4 e^ { \frac{5}{2 } t}-\frac{7}{32 } m^2 e^ { \frac{1}{2}t } \\ & & + \frac{1}{16 } m^2 e^ { \frac{3}{2 } t}+\frac{5}{32 } m^2 e^ { \frac{5}{2 } t}+\frac{15}{256 } e^ { \frac{1}{2}t } -\frac{3}{128 } e^ { \frac{3}{2 } t } -\frac{9}{256 } e^ { \frac{5}{2 } t } \bigg ] \\ & & + 4 e ^{-t } \bigg [ \frac{1}{16 } m^4 e^{-\frac{1}{2}t } -\frac{1}{8 } m^4 e^{\frac{1}{2}t}+\frac{1}{16 } m^4 e^ { \frac{3}{2 } t}\\ & & + \frac{7}{32 } m^2 e^{-\frac{1}{2}t } -\frac{3}{16 } m^2 e^{\frac{1}{2}t}+\frac{7}{32 } m^2 e^ { \frac{3}{2 } t } -\frac{15}{256 } e^{-\frac{1}{2}t } -\frac{9}{128 } e^{\frac{1}{2}t } -\frac{15}{256 } e^ { \frac{3}{2 } t } \bigg ] \\ & = & - \left ( \frac{1}{4}- m^2\right ) e ^{\frac{t}{2}}\,.\end{aligned}\ ] ] this proves the last statement .
proposition is proven .
the proof is straightforward ; we just substitute the function @xmath236 ( [ 1.21 ] ) into equation ( [ 1.20 ] ) and then check the initial conditions . in order to make proof more transparent
we split it into three independent cases .
* case of ( @xmath237 ) * in this case set @xmath238 , @xmath239 .
henceforth we suppress the subindex @xmath240 of @xmath241 .
we have @xmath242 and consider the derivative @xmath243 according to the choice ( [ 1.23 ] ) of the function @xmath244 , we have @xmath245 now we take into account ( [ 1.23 ] ) , that is @xmath246 , and the second statement @xmath247 of proposition [ p2.13 ] .
hence we have @xmath248 due to proposition [ p2.13 ] and proposition [ p2.14 ] we can write @xmath249 this competes the proof of the case ( @xmath250 ) .
similarly , we consider the next case .
* case of ( @xmath251 ) * in this case we set @xmath238 , @xmath252 , then ( [ 1.21 ] ) reads @xmath253 from now on we skip subindex @xmath254 of @xmath255 . for the first and second - oder derivatives we have @xmath256 and @xmath257 respectively , while @xmath258 consider the last integral ; using twice integration by parts and the choice of function @xmath259 , we obtain @xmath260 since @xmath261 , we obtain @xmath262 hence @xmath263 \\ & & + 2\int _ { 0}^{\phi ( t ) } v _ { } ( x , s ) \bigg [ k_{0\,tt}(s , t;m)\ , ds -e ^{-2 t } k_{0\,ss}(s , t;m)\ , ds - m^2 k_0(s , t;m)\bigg ] \ , ds\,.\end{aligned}\ ] ] then we use proposition [ p2.15 ] , which completes the proof of the case of @xmath264
. * case of ( f ) * similarly , we set @xmath265 , @xmath266 , then ( [ 1.21 ] ) reads @xmath267 for the first and second - order derivatives we obtain @xmath268 and @xmath269 respectively . on the other hand , due to the definition of the kernel @xmath270 , lemma [ l2.2 ] , and ( [ 2.1 ] ) , we obtain @xmath271 hence , since @xmath272 , we have @xmath273 then , taking into account ( [ 1.22 ] ) , we have @xmath274 the definition of @xmath275 , ( [ 1.21 ] ) , and ( [ 3.10 ] ) imply @xmath276 hence @xmath277 on the other hand , the integration by parts leads to @xmath278 then we use proposition [ p2.3 ] .
consequently , @xmath279 thus , @xmath280 now we apply theorem [ t2.12 ] to the last integral and obtain @xmath281 proposition [ p2.9 ] completes the proof of this case .
it is easy to check the initial conditions .
theorem [ t1.2 ] is proven .
@xmath142 k. yagdjian , semilinear hyperbolic equations in curved spacetime .
fourier analysis , pseudo - differential operators , time - frequency analysis and partial differential equations .
series : trends in mathematics .
birkhuser mathematics , 391415 , 2014 . | in this paper we describe the integral transform that allows to write solutions of one partial differential equation via solution of another one .
this transform was suggested by the author in the case when the last equation is a wave equation , and then it was used to investigate several well - known equations such as tricomi - type equation , the klein - gordon equation in the de sitter and einstein - de sitter spacetimes .
a generalization given in this paper allows us to consider also the klein - gordon equations with coefficients depending on the spatial variables .
+ _ keywords : _ klein - gordon equation ; curved spacetime ; representation of solution 8.5 in -0.15
cm -1.5 cm 6.5 in [ section ] [ theorem]lemma [ theorem]definition [ theorem]corollary [ theorem]remark [ theorem]proposition [ theorem]example [ theorem]assumption _ department of mathematics + university of texas - pan american + 1201 w. university drive + edinburg , tx 78539 usa _ |
the study of topological or vacuum defects is of importance in many areas of contemporary physics . in high energy physics ,
a defect will generically occur during a symmetry breaking process where different parts of a medium choose different vacuum energy configurations , and the non - compatibility of these different vacua forces a sheet , line , or point of energy where these non - compatible vacua meet .
the relevant vacuum order parameter then becomes indeterminate this is the defect .
a defect may be topological @xcite , in that it is the topology of the vacuum that simultaneously allows formation , and prevents dissipation , of these objects but other types of defect are also possible .
for instance , a defect may be stable dynamically ( i.e. classically , due to energy considerations ) but not topologically , as it happens for semilocal @xcite or electroweak @xcite defects .
a defect can even be ` topological ' and unstable , as in the case of textures @xcite , but nonetheless of physical importance .
in cosmology , there has been much speculation that topological defects ( stable or otherwise ) might have played an important role in structure formation @xcite . in general
, there are two main concerns when considering the cosmological effects of topological defects : their gravitational effects and their dynamics .
any theory concerning galaxy formation must be able to allow or constrain the presence of strongly self - gravitating objects .
but the dynamics of the defects are in fact even more important , for it is the dynamics that determine to a large extent the shapes of the gravitating defects .
for instance , if cosmic strings did not intercommute , any network would rapidly become tangled and would not obey a scaling law ; such a configuration would be in conflict with the universe we see around us today .
even with intercommutation @xcite , strings which are strongly rigid and hence straight will have different gravitational effects to those that are very crinkly @xcite .
the dynamical behaviour of a defect is generally assumed to be approximated by an effective action , a description which models the rather large numbers of degrees of freedom of the full field theory by the smaller number of degrees of freedom based on the position of the core of the defect .
attempts to derive effective actions or equations of motion for topological defects have commonly focussed on the strong coupling limit , meaning that of large values of the coupling coefficient @xmath0 of the relevant higgs field . in this limit , the defect becomes infinitesimally thin and effectively decouples from the other ( infinitely massive ) particles in the field theory .
the study of the effective motion of topological defects has been extended @xcite away from the limit @xmath1 to cases for which the thickness is small but not exactly zero .
the resultant effective action generically contains a ` zero - thickness ' term proportional to the area of the defect @xcite , and extrinsic curvature terms which appear at quadratic order in the thickness @xcite .
the exact form of these second order terms is dependent on whether an ` off - shell ' or ` on - shell ' approach has been used to derive the effective action as was explained using the example of the domain wall in @xcite , and one finds that in a self - consistent order by order solution of the equations of motion , the only quadratic correction appearing is due to the geometry of the defect worldsurface , and is proportional to its intrinsic ricci curvature @xcite . for the domain wall , this term gives corrections to the motion , however for the string this term is a topological invariant
proportional to the euler character of the worldsheet and hence does not give any correction to the nambu equations of motion .
the leading order corrections to the motion of a cosmic string were shown by one of us , @xcite , to appear at quartic order in the string width . in this paper , a systematic expansion of the geometry and field equations clarified an earlier discrepancy concerning second order ` twist ' terms @xcite , and derived the fourth order action of the string , which is @xmath2 , \label{anderson}\ ] ] where @xmath3 , @xmath4 are the ricci and extrinsic curvatures of the worldsheet ( to be defined in the next section ) , and the @xmath5 are numerical coefficients depending on the field theory modelling the vortex ( to be defined in section [ sec : correct ] for the abelian - higgs model ) . unlike the domain wall ,
whose background solution , and extrinsic curvature corrections can be expressed analytically @xcite , the prototypical cosmic string solution , the nielsen olesen vortex @xcite , does not have a closed analytic form , hence these coefficients must be evaluated numerically .
as soon as one calculates corrections to the nambu action , it becomes of general concern whether or not these corrections cause the motion of the defect to be ` rigid ' or ` antirigid . '
the implications of extrinsic curvature terms for string motion have been well - explored @xcite , however , it is not always easy to decide _ a priori _ ( especially for the fourth order terms ) whether the strings will be rigid or not , or indeed what one means by rigid @xcite .
in this paper we address these issues .
we determine the numerical values of the @xmath5 for the abelian - higgs model by solving the perturbed field equations for the non - flat worldsheet .
we then derive the fourth order equations of motion from anderson s action and calculate corrections to three ` test case ' trajectories : the circular loop , the travelling wave , and the helical breather .
these three solutions display different characteristics which should be mirrored in the corrections to the nambu motion if the rigidity of the string is to be determined .
the loop collapses to a point @xcite , and rigidity would be indicated by a retardation of this collapse .
a travelling wave on the other hand has been shown to be an exact solution of the full field theory @xcite , and should not exhibit any corrections if our approach is to be trusted . the helical breather
( see e.g. @xcite ) is a time dependent solution which is never singular .
rigidity for this trajectory is more subtle , since the helical breather is never singular , however , we could call the string rigid if the tendency of the correction is to lower the magnitude of the scalar curvature of the worldsheet . as we will see , it is rather difficult to give an intuitive criterion for rigidity .
the layout of the paper is as follows . in the next section we review the formalism for the derivation of the effective action . in section [ sec : correct ] we rederive anderson s action , and present new numerical results evaluating the coefficients appearing in the action . in section [ sec : eom ] we derive the equations of motion of the fourth order string , and in section [ sec : traj ] we calculate corrections to three test trajectories . in section [ sec : rigid ] we discuss the question of rigidity and conclude .
in this section we review the formalism required for the derivation of the string effective action .
this formalism is largely based on the contents of @xcite .
the problem of building the effective action is to reduce some four - dimensional field theoretic action integral @xmath6 to some two - dimensional worldsheet integral @xmath7 we follow the canonical approach by setting up a coordinate system based on the worldsheet , and expanding the action and equations of motion around this worldsheet . by `
expand ' we mean that we do not expect in one go to be able to solve the full equations of motion and integrate out ( otherwise why find an effective action ) but that we will be able to express the equations of motion for the system in terms of an expansion in @xmath8 , where @xmath9 is the radius of the string , and @xmath10 a typical scale of the extrinsic curvature of the worldsheet .
we then solve the equations of motion in the directions perpendicular to the worldsheet , and integrate out the four - dimensional action ( [ actfour ] ) over these degrees of freedom to get an action of the form ( [ actwo ] ) .
we will look at the particular case of a u(1 ) local string in flat spacetime ( signature @xmath11 ) .
this is a vortex solution of the abelian - higgs model @xmath12 where @xmath13 is the usual gauge covariant derivative , and @xmath14 the field strength associated with @xmath15 .
we rewrite the field content of this model in the usual way : [ changevar ] @xmath16 , \end{aligned}\ ] ] so that the vacuum is characterised by @xmath17 . in terms of these new variables ,
the lagrangian becomes @xmath18 where @xmath19 is now the field strength of @xmath20 . by making a compensating gauge transformation in @xmath15 ,
we have absorbed the unphysical gauge degree of freedom of the theory . for future reference , here are the equations of motion for the new fields : [ eomxp ] @xmath21 these equations are known to admit vortex solutions , the best known being that due to nielsen and olesen @xcite a solution corresponding to an infinite straight static string .
a vortex solution is characterised by the fact that the scalar field phase has a non - zero winding number around an axis corresponding to the zeroes of the higgs field ( x=0 ) .
if we choose the string axis to be aligned with the @xmath22-axis , then a gauge can be chosen in which the vortex solution takes the form @xmath23 \nabla_\mu \theta\ ] ] in cylindrical polar coordinates , where @xmath24 and @xmath25 satisfy the ( generically ) second order coupled differential equations : [ no ] @xmath26 and we have introduced the bogomolnyi parameter @xmath27 .
@xmath28 is the global string , and @xmath29 is the critical ( supersymmetric ) string for which the equations of motion factor to two coupled first order equations : @xmath30 intuitively we expect the solution for a general worldsheet to be very close to the nielsen
olesen solution in suitable coordinates , and that we might be able to expand the fields , and therefore the action , around this approximate solution .
let us now try to make this idea more concrete .
first we choose a worldsheet - based coordinate system .
( note this will only be valid within the radii of curvature of the worldsheet . )
we start by coordinatising the worldsheet by @xmath31 and @xmath32 as usual .
since we are in flat space , at each point on the worldsheet we have a well defined orthogonal flat two - plane , and such orthogonal planes do not intersect within the radii of curvature of the worldsheet .
thus we can thicken the worldsheet to a worldblanket by defining @xmath31 and @xmath32 to be constant on such orthogonal planes .
since the worldsheet @xmath33 has codimension 2 , it has two associated families of unit normals @xmath34 .
we choose a cartesian parametrisation of the orthogonal planes such that @xmath35 .
thus @xmath36 define a coordinate system in the vicinity of the worldsheet .
in general it will not be possible to choose this system to be globally orthogonal , however , we will assume that the normal fields have been chosen so that the departure from orthogonality is minimal , i.e. of order of the extrinsic curvature of the worldsheet . because we no longer have a flat coordinate system ,
the connection will no longer be trivial , although the curvature components should still vanish . in order to examine the form of these relations , we use a gauss
codazzi approach @xcite ( see also @xcite for more of a physicist s approach ) . for shorthand
we denote by @xmath37 ( @xmath38 ) the coordinates @xmath39 , the worldsheet by @xmath33 and the minkowski spacetime by @xmath40 .
the _ first fundamental form _ @xmath41 of @xmath33 is defined as @xmath42 the tensor @xmath41 acts as a projector onto @xmath33 , but is still a tensor in the 4-dimensional spacetime . for the metric _ intrinsic _ to the worldsheet
, one uses the familiar @xmath43 where @xmath44 are the spacetime coordinates of the submanifold @xmath33 .
this gives an interpretation of @xmath33 as both a worldsheet in spacetime and a two - dimensional manifold with its own intrinsic geometry .
one can define the intrinsic curvature of @xmath33 in the standard way , and the extrinsic curvatures , or _
second fundamental forms _ , by @xmath45 these extrinsic curvatures @xmath4 measure how @xmath33 curves in @xmath40 .
since the codimension of the worldsheet is greater than one , we also have a non - trivial _ normal fundamental form _
@xmath46 where @xmath47 is the alternating symbol on two indices .
this represents how near to orthogonality the worldblanket coordinate system is ; it measures the rotation of the @xmath48 in their own planes as one moves around the worldsheet .
note that , despite lying tangent to @xmath33 , @xmath49 is a gauge dependent object , depending on the choice of the normal fields
an so(2 ) gauge group .
in fact , @xmath49 is the connection on the normal bundle of @xmath33 .
now , we may write @xmath50 where the symmetrization is understood to be only acting on indices of the same type , so that the ` @xmath51 ' above does not participate in any symmetrization . this now indicates an alternative definition of the second fundamental forms : @xmath52 where @xmath53 denotes the lie derivative with respect to @xmath48 .
we may now use the riemann identity in flat space to derive the following relation for the extrinsic curvatures : @xmath54 therefore , contracting with @xmath55 gives @xmath56 in addition , note that @xmath57 and @xmath58 the system of equations ( [ altk][lien ] ) give the ` equations of motion ' for the geometry of the system , which together with ( [ eomxp ] ) form the full equations of motion for the string . in order to extract a low energy effective action , we will need to expand these quantities and equations of motion in terms of the thickness of the string , @xmath9 . to do this
systematically we define the dimensionless parameter @xmath59 by @xmath60 where @xmath61 represents a typical radius of curvature of the worldsheet .
we then define the _ zero thickness limit _ to mean @xmath62 with @xmath63 , the energy per unit length of the string , fixed .
note that this may not necessarily mean that the string width _ is _ zero , since instead the extrinsic curvature could be zero ( a flat worldsheet ) .
however , if @xmath64 , this limit corresponds to the conventional @xmath65 limit .
we now redefine coordinates so that the worldsheet has thickness and curvature of order unity by setting [ rescale ] @xmath66 and redefining for consistency the following variables [ redefine ] @xmath67 the worldsheet geometry equations ( [ altk][liebeta ] ) become [ unitgeom ] @xmath68 \\ { \cal l}_j k_{_i\mu\nu } & = & \epsilon k_{_i\sigma(\mu } k_{_j\nu)}^\sigma \\ { \cal l}_j k_{_i } & = & - \epsilon k_{_i\mu\nu}k_{_j}^{\mu\nu } \\ { \cal l}_j \omega_\mu & = & 0 , \end{aligned}\ ] ] and the field equations ( [ eomxp ] ) become : @xmath69 \bigg\ } \nonumber \\ & & = { \hat p}_\mu { \hat p}_\nu g^{\mu\nu}x - { \frac12 } x(x^2 - 1 ) \label{xorder}\end{aligned}\ ] ] @xmath70 \bigg\ } + \nonumber \\ & & \epsilon \ , \partial_a \bigg\ { \sqrt{-g } \bigg [ g^{ak}g^{jl } { \hat f}_{kl } + ( g^{ak}g^{jb } - g^{ab}g^{jk } ) ( \partial_k { \hat p}_b - \epsilon \partial_b { \hat p}_k ) + \epsilon g^{ab } g^{jc } { \hat f}_{bc } \bigg ] \bigg\ } \bbox{\bigg ) } \nonumber \\ & & = -\beta x^2 ( g^{ij } { \hat p}_i + g^{ja}{\hat p}_a ) \label{piorder}\end{aligned}\ ] ] @xmath71 \bigg\ } + \nonumber \\ & & \epsilon \ , \partial_b \bigg\ { \sqrt{-g } \bigg [ g^{bk}g^{al } { \hat f}_{kl } + ( g^{bc}g^{al } - g^{bl}g^{ac})(\epsilon\partial_c{\hat p}_l -\partial_l{\hat p}_c ) + \epsilon g^{bc}g^{ad } { \hat f}_{cd } \bigg ] \bigg\ } \bbox{\biggr ) } \nonumber \\ & & = -\beta x^2 ( g^{ja}{\hat p}_j + g^{ab}{\hat p}_b ) .
\label{paorder}\end{aligned}\ ] ] here , @xmath72 is the tensor @xmath73 defined in terms of the rescaled variables . finally , we also need the explicit expansion of the lagrangian @xmath74 + \nonumber \\ & & \hspace*{8 mm } 2 ( \partial_i { \hat p}_a - \epsilon \partial_a { \hat p}_i ) ( \partial_j { \hat p}_b - \epsilon \partial_b { \hat p}_j ) ( g^{ij}g^{ab } - g^{ib } g^{ja } ) + \nonumber \\ & & \hspace*{8 mm } \epsilon { \hat f}_{ac } \bigg [ 4(\partial_i { \hat p}_b - \epsilon \partial_b { \hat p}_i ) g^{ia } g^{bc } + \epsilon { \hat f}_{bd } g^{ab } g^{cd } \bigg ] \bigg\ } - { \frac14 } ( x^2 - 1)^2 \bbox{\bigg)}. \label{lorder}\end{aligned}\ ] ] this now allows us to expand rigorously in powers of @xmath59 .
note that so far we have made no assumptions about any of the fields or their dependence on the coordinates .
we have simply rewritten the equations of motion , scaling with respect to the physical dimensionful quantities in the problem , leaving the equations of motion in terms of the dimensionless parameter @xmath59 .
the procedure is as follows .
we first consider the @xmath75 limit , with the metric and fundamental forms taking their geometrically defined values on the worldsheet .
since @xmath75 is also the flat worldsheet limit , we expect that the fields will take the nielsen
olesen form , which indeed turns out to be the case . in the next section
we will go to higher orders in @xmath59 , deriving the corrections to the geometry and fields , and hence the corresponding effective action .
however , for the moment we conclude this section by proving that the nielsen
olesen solution is in fact the leading order behaviour for the vortex fields around the worldsheet . to zeroth order @xmath76
hence all geometrical quantities take their worldsheet values , @xmath77 etc .
hence @xmath78 and thus the equations of motion for @xmath79 and @xmath80 are [ xpzero ] @xmath81 these are solved by the nielsen
olesen solution , which is plotted in figure [ fig : no ] against @xmath82 for @xmath29 : @xmath83 substitution of this solution back into ( [ actfour ] ) yields the familiar nambu action .
it is the corrections to this action we are interested in .
we now derive the leading order corrections to the nambu action . in order to calculate quantities away from the worldsheet , such as the metric
, we must perform a taylor expansion off the worldsheet @xmath84 however , in calculating the lie derivatives of the metric and its determinant some fortuitous cancellations occur . first note that @xmath85 and also that @xmath86 which follows from a trace identity for the @xmath87 matrix @xmath88 @xcite . hence to all orders , the metric and the volume jacobian are given by @xmath89.\end{aligned}\ ] ] this gives all the information on the geometrical contributions to the action .
now let us turn to the field theoretic contributions .
first note that since @xmath90 vanishes by the equations of motion , first order field corrections contribute at second order in the action , and second order corrections at fourth order .
to first order we may read off the equations for the corrections to the fields as [ first ] @xmath91 the first two equations do not apparently lend themselves to a straightforward solution , however , we note that the equation of motion for the worldsheet obtained by varying the nambu action is @xmath92 which would indicate that the ` driving ' terms in ( [ first_a ] , [ first_b ] ) vanish and hence the appropriate solution is @xmath93 . indeed , @xcite demonstrated that unless the trace of the extrinsic curvature vanished on the defect , the first order perturbation equations have no solution that is regular and bounded .
thus in fact the appropriate solutions to the first two equations of ( [ first ] ) are @xmath94 and @xmath95 .
the last equation of ( [ first ] ) is solved by @xcite @xmath96 the presence of this term , as pointed out in @xcite , guarantees the gauge invariance of the effective action to worldsheet so(2 ) gauge transformations , for on substituting in the form of the fields , this correction to @xmath97 exactly cancels the ` twist ' terms of maeda and turok , and we arrive at the second order result [ firstact ] @xmath98 where [ fcoeff ] @xmath99 for the fourth order action we need the second order corrections to the fields , the volume factor already being exact to all orders . from ( [ xorder][paorder ] )
we may read off the equations for these as [ xpsecond ] @xmath100 to simplify these , and to remove the explicit @xmath88-dependence we will decompose in cylindrical harmonics by setting [ secondh ] @xmath101 where @xmath102 this gives two sets of coupled second order differential equations [ zerohh ] @xmath103 and [ secondeq ] @xmath104 there are no additional corrections to @xmath97 at this level . in the critical case @xmath29
, the above equations reduce to two sets of first order ode s : [ eq : crit ] @xmath105 and [ eq : crit_2 ] @xmath106 note that just as ( [ secondeq ] ) is really two coupled second order ode s , ( [ eq : crit_2 ] ) is really two coupled first order ode s .
the fact that ( [ zerohh ] , [ secondeq ] ) reduce to such a first order form for @xmath29 is reassuring , since we do not expect to destroy supersymmetry simply by bending the string ! substituting ( [ secondh ] ) into the action , and after some algebra , one arrives at anderson s result : [ fourth ] @xmath107 \bigg\ } \\ & = & - \mu \int d^2 \sigma \sqrt{-\gamma } \bigg\ { 1 - r_s^2 { \alpha_1\over\mu } { \cal k}_{_i\mu\nu}^2 \hspace{1 mm } + r_s^4 \bigg [ { \alpha_2\over\mu } ( { \cal k}_{_i\mu\nu}^2)^2 + { \alpha_3\over\mu}{\cal k}_{_i\mu\nu}{\cal k}_{_j}^{\mu\nu } { \cal k}_{_i\lambda\rho } { \cal k}_{_j}^{\lambda\rho}\bigg ] \bigg\ } , \end{aligned}\ ] ] where [ scoeff ] @xmath108 \\ \alpha_3 & = & { \pi\eta^2\over 4 } \int_0^\infty dr \bigg [ 2r^2x_{_0}'{\bar g } + { 2\hat{p}_{_0}'\over\beta}({\bar q_\phi } + r q_r ) - { 4r\hat{p}_{_0}^2 \over\beta } \bigg ] .
\end{aligned}\ ] ] if @xmath29 , it can be seen from ( [ eq : crit ] , [ eq : crit_2 ] ) that @xmath109 . to determine @xmath110 and @xmath111 , we need to solve equations ( [ zerohh ] ) and ( [ secondeq ] ) .
this was done numerically by relaxation methods , using nag routine d02gaf @xcite .
sample solutions for both sets of equations for the critical case @xmath29 are plotted in figures [ fig:21 ] and [ fig:22 ] .
the first set of equations can be readily implemented and was solved by requiring that @xmath112 . before solving the second set of equations , we expressed @xmath113 from ( [ secondeq_c ] ) as : @xmath114 and eliminated it from the equations , which were then solved by asking that @xmath115 . the solutions obtained clearly behave at the origin as : @xmath116 which can be checked to satisfy the asymptotic counterparts of ( [ zerohh ] , [ secondeq ] ) .
we have computed the coefficients @xmath117 and @xmath111 for a range of values of @xmath118 ; the results are shown in figures [ fig : mualpha1][fig : alpha3 ] and in table [ table ] .
the key points to note are that for the range of @xmath118 for which we computed the coefficients ( @xmath119 ) @xmath120 and @xmath121 ; however , @xmath122 is positive ( negative ) for @xmath123 ( @xmath124 ) and @xmath110 is positive for @xmath125 and negative for @xmath126 .
in order to derive the equations of motion for the string we must express the action ( [ fourth ] ) in terms of the worldsheet coordinates , with respect to which we are varying the action . to do this , we note that @xmath127 hence , defining @xmath128 we see that [ trems ] @xmath129 now , the connection on the worldsheet is given by @xmath130 and the curvature ( either directly or from the gauss codazzi equations ) is @xmath131 we may therefore , using the identity @xmath132 and the symmetries of @xmath133 , infer the following useful relations : [ userel ] @xmath134 we can now rewrite the action and its constituent variations as @xmath135 \bigg\}\ ] ] and [ vars ] @xmath136 [ using the fact that @xmath137 is parallel to the worldsheet whereas @xmath138 is perpendicular , and ( [ userel ] ) . ] noting that @xmath139 , the euler characteristic which is a topological invariant , or from the above equations , we see that the @xmath140 term will not contribute to the equations of motion .
the @xmath141 term , on the contrary , gives @xmath142 where we used the riemann identity @xmath143 which implies @xmath144 at each step .
the @xmath145 term gives @xmath146 but @xmath147 hence the full equations of motion for the system are : @xmath148 it is possible to express these equations in terms of the extrinsic curvatures . using ( [ malcone ] ) and
@xmath149 it follows that @xmath150 contracting the equation of motion with @xmath151 and with @xmath152 gives equations of motion normal and parallel to @xmath33 : @xmath153 note that this latter equation ( [ eompar ] ) is an identity for the unperturbed worldsheet .
we have verified that it does indeed hold using the lightcone gauge .
now , @xmath154 which implies that the equations of motion also read : @xmath155
we now wish to derive the corrections to the motion of three test trajectories : the circular loop , the travelling wave , and the helical breather .
we have chosen these three examples because they provide three different ways in which to observe the rigidity or otherwise of the string .
the loop trajectory is given by @xmath156 and collapses to a point after a time period @xmath157 , where @xmath158 is the length of the closed string .
the extrinsic curvature invariants become singular at this point , hence rigidity would be indicated by a retardation of the collapse or a positive correction to the amplitude of the loop .
a travelling wave on the other hand is a variant of the flat worldsheet where a deformation of arbitrary size and form is introduced , the only constraint being that the deformation is a function of only one of the lightcone coordinates , @xmath159 : @xmath160 this has been shown to be a solution of the full field theory @xcite , hence no correction should be found for this trajectory . the helical breather is a time - dependent solution which may be written as @xmath161 the limit @xmath162 corresponds to the flat worldsheet and the limit @xmath163 to the collapsing circular loop . for intermediate @xmath164
the trajectory is never singular , and the extrinsic curvature peaks at approximately @xmath165 .
rigidity would be indicated by a preference for lower extrinsic curvature and hence a negative correction to the amplitude of oscillation . as a first example of the effects of curvature terms on the motion of freely - moving cosmic strings , we consider the collapse of a circular loop . in cartesian
coordinates , the position of this string is given by @xmath166 and the normals to the worldsheet are [ loop : normal ] @xmath167 in this case , one obtains @xmath168 , and @xmath169 the equation of motion to zeroth order is @xmath170 the general solution to this equation is @xmath171/k)$ ] . choosing @xmath172 as initial conditions one obtains the canonical form of the loop trajectory ( [ ltraj ] ) .
note that for this choice of loop length @xmath173 and we can use @xmath88 or @xmath174 interchangeably .
we now wish to find the corrected solution to order @xmath175 . to do this , we use equation ( [ eomextr ] ) ( technically its unrescaled counterpart ) .
since @xmath176 and @xmath177 vanish , the rhs of ( [ eomextr ] ) considerably simplifies to @xmath178 ( note that we have dropped the subscript ` 2 ' on the extrinsic curvature . ) the left - hand side is obtained by varying the trace of @xmath179 from ( [ loop : zeroth ] ) , whereby we obtain the equation for @xmath180 as @xmath181 the solution to this , with initial conditions @xmath182 is @xmath183 thus , the sign of @xmath184 determines whether the string is rigid or antirigid .
this ` loop rigidity parameter ' is plotted against @xmath118 on figure [ fig : rigidity ] , and its negativity means that the loop is antirigid and tends to collapse faster than in the nambu approximation .
this is illustrated on figure [ fig : collapse ] , where we compare the zero - order solution @xmath185 with the corrected solution @xmath186 .
note that the approximation breaks down when @xmath187 , i.e. when @xmath188 .
for example , in the illustration of collapse in figure [ fig : collapse ] , @xmath189 is rather large ; we would hope that our approximation would be valid until the radii of curvature of the worldsheet became close to 1/10 , let us say twice the radius of the string : 1/5 .
inputting @xmath190 gives @xmath191 , which does indeed correspond to the point at which the solutions start to significantly differ in figure [ fig : collapse ] .
now consider a travelling wave , whose position and normals are given by : @xmath192 where a prime denotes differentiation with respect to the argument , namely @xmath193 .
writing @xmath194 one has : @xmath195 it is then straightforward to see that all terms on the right - hand side of ( [ eomextr ] ) vanish separately , as do the traces of the extrinsic curvatures : the travelling wave is an exact solution to ( at least ) fourth order .
we now consider the string given by the following position functions : @xmath196 this string is helical with breathing @xmath164 : the limits @xmath197 and @xmath198 represent a collapsing loop and a straight string , respectively .
[ note how ( [ helical : position][helical : normal2 ] ) reduce to ( [ loop : position][loop : normal ] ) , and how all pertinent quantities in the equations of motion are obtained from these expressions . ] with this choice of normals , the fundamental forms are : @xmath199 hence the equations of motion to zeroth order become @xmath200 this equation admits for general solution @xmath201 so that , choosing again the initial conditions @xmath172 and calling @xmath202 we have @xmath203 the right - hand side of the corrected equation of motion becomes then @xmath204 where [ eq : betas ] @xmath205 \\ \beta_2 & = & 7 ( \alpha_2 + \alpha_3 ) + ( 38 \alpha_2 + 22 \alpha_3 ) q^2 + ( 7 \alpha_2 - 3 \alpha_3 ) q^4
\\ \beta_3 & = & 5 q^2 \left [ ( 7 \alpha_2 + 5 \alpha_3 ) + ( 7 \alpha_2 + 2 \alpha_3 ) q^2 \right ] \\
\beta_4 & = & 15 q^4 ( 2 \alpha_2 + \alpha_3 ) , \end{aligned}\ ] ] and the left - hand side is obtained as before by varying the trace of @xmath206 : @xmath207\delta z. \label{helical : deltak}\ ] ] finally , the corrected equations of motion are : @xmath208 although it is possible to find an exact solution to this equation ( see appendix for details ) , it is more instructive to consider the quasiflat limit @xmath198 . in this case ,
equation ( [ helical : fulleom ] ) becomes @xmath209,\ ] ] where @xmath210 is defined by @xmath211 the solution @xmath212 satisfying @xmath213 is then found to be @xmath214\ ] ] the corrected trajectory can then be written : @xmath215 \cos \left\ { \left [ 1 + \frac{4 \epsilon^4\delta^2}% { \mu } ( 2 \alpha_2 + 3 \alpha_3 ) \right ] \tau \right\ } + \frac{\epsilon^4 \delta^3}{\mu } ( 2 \alpha_2 + \alpha_3 ) \cos ( 3\tau ) + \mbox{o}(\epsilon^8).\ ] ] the effect of the correction is threefold : first , it alters the frequency of the motion , @xmath216 \tau$ ] ; since @xmath217 this has the effect of reducing the frequency a tendency we would be tempted to call rigid . secondly , the amplitude of the oscillation is altered by a factor @xmath218 .
this could be either a amplification or reduction , depending on whether @xmath124 or @xmath123 .
finally , a higher frequency oscillation is superposed on the motion for @xmath219 .
if , for simplicity , we take @xmath29 , so @xmath220 , we see that @xmath221\ ] ] i.e. the only effect of the correction is to reduce the frequency of oscillation of the breather , which would seem to be unambiguously rigid .
however , we now observe a curious property : suppose instead we consider initialising the correction at the instant of maximal velocity : @xmath222 , we find @xmath223 \sin \left\ { \left [ 1 + \frac{4 \epsilon^4 \delta^2}{\mu } ( 2 \alpha_2 + 3 \alpha_3 ) \right ] \tau ' \right\ } - \frac{\epsilon^4 \delta^3}{\mu } ( 2 \alpha_2 + \alpha_3 ) \sin(3 \tau ' ) + \mbox{o}(\epsilon^8)\ ] ] ( where @xmath224 ) .
now note that while the frequency of oscillation is decreased by the same amount , and the higher frequency term is still the same , the amplitude is now uniformly increased for all @xmath118 . if , as before , we take @xmath29 , we now find @xmath225,\ ] ] in other words an _ increase _ in the amplitude of oscillation accompanies a similar decrease in its frequency .
one can not immediately see from this solution whether or not the behaviour is rigid , however , an analysis of the ricci curvature near @xmath226 shows that it is in fact increased a behaviour consistent with antirigidity ! a calculation of the dependence of the curvature on general @xmath164 is presented in the appendix . the algebra is more complicated but the results are the same : all helical breathers display both rigid and antirigid characteristics .
the results of our helical breather calculations therefore appear rather ambiguous .
if we wish to identify rigidity by the behaviour of a corrected trajectory whether it increases or decreases curvature we are forced to calculate the effect on the curvature ( see appendix ) and then the results appear to depend on the initial conditions .
what this shows is that the ` decrease in curvature ' criterion for rigidity is too nave to be reliably applied in all situations .
we now want to determine whether a string can be labelled ` rigid ' or ` antirigid ' . to do this
, we use an argument based in that of polyakov @xcite , which consists in determining how the action varies under a rescaling of the spacetime coordinates @xmath227 .
such transformations alter the scale of crinkles of the worldsheet and magnify or reduce small - scale structure , hence rigidity would be indicated by a extremum of the energy or the action with respect to the rescaling parameter @xmath0 , as illustrated in figure [ fig : rescale ] .
out starting point is the action ( [ fourth ] ) , which can be written @xmath228 where the matrix @xmath229 is defined as @xmath230 as @xmath231 , this can be expressed as @xmath232,\ ] ] with @xmath233 the area of the worldsheet for the range of @xmath234 being integrated over and @xmath235 proportional to the euler character of @xmath33 . also , @xmath236 we now rescale @xmath227 , so that for @xmath237 the worldsheet is expanded and for @xmath238 it is shrunk . then , @xmath239 and thus @xmath240 , @xmath241 .
the shape of the curve @xmath242 now depends explicitly on the integrals @xmath243 , since @xmath244 is rescaled as @xmath245.\ ] ] we know that @xmath246 , and clearly @xmath247 , so in order to determine the shape of @xmath242 we only need to determine the sign of @xmath248 . for these purposes
, we can work in the conformal gauge , @xmath249 , and we find @xmath250 if we impose the nambu equations of motion , @xmath251 , this determinant is strictly negative .
hence , since @xmath252 is positive , we see that @xmath242 is unbounded , because the the coefficients multiplying @xmath253 and @xmath254 have opposite signs .
we must therefore conclude that the string is antirigid .
this does not mean that _ all _ the trajectories of the string exhibit antirigidity , but rather that it is impossible for all trajectories to be rigid .
let us illustrate this by considering the trajectories of the previous section .
for the collapsing loop , @xmath255 , and as we noticed , @xmath256 determines alone the shape of @xmath242 .
this simplification came from the fact that the loop is flat , and therefore has only one nonvanishing extrinsic curvature . for
the travelling wave , @xmath257 , which is consistent with our observation of no corrections to this trajectory . in the case of the helical breather , both @xmath258 and @xmath248 are non - zero , so we do not expect results to depend on @xmath184 . even though the action is unstable to the scaling of the worldsheet
, what is happening with the helical breather is that the correction does not always have a non - zero projection on this unstable mode .
however , we would expect a general correction of the breather to exhibit an instability . to sum up :
we have reviewed the derivation of the effective action for a u(1 ) local cosmic string to fourth order in the ratio of string width to worldsheet curvature .
we presented numerical results calculating the coefficients of these fourth order terms .
we then derived the equations of motion for the string to fourth order , and calculated corrections to a sample of well - known trajectories .
we have given a general argument for antirigidity of the cosmic string to fourth order , however by reference to our examples have shown that not all trajectories need behave in an ` antirigid ' fashion rigidity it appears is rather like a theorem , it may work in special cases , but one needs only find a single counterexample to disprove it .
the authors would like to thank lyndon woodward for useful discussions and bernard piette for help with numerical aspects of this work .
rg was supported by the royal society and fb was partially supported by a grant from the board of the swiss federal institutes of technology .
recall from section [ sec : breather ] that the helical breather solution to the nambu action is @xmath259 [ where @xmath210 is defined in ( [ eq : delta ] ) ] and that the equation of motion for @xmath260 is ( [ helical : fulleom ] ) : @xmath261 this can be solved using the method of variation of parameters , giving ( after a long and tedious calculation ) : @xmath262 \nonumber \\ & & + \lambda_0 \sin(\tau)\tan^{-1 } \big ( q \tan(\tau ) \big ) + \sin^2(\tau)\cos(\tau ) \sum_{n=1}^{6 } \lambda_n \omega^{-n}.\end{aligned}\ ] ] here , @xmath263 , the @xmath118 s were defined in ( [ eq : betas ] ) and the @xmath0 s are : @xmath264 , \\
\lambda_1 & = & \frac{\delta}{720 \mu } \bigg [ 1920 \big(3 - 1/q^2\big ) \beta_1 - 360 \big ( 15 - 4/q^2 - 3/q^4 \big ) \beta_2 + 48 \big ( 105 - 25/q^2 - 17/q^4 - 15/q^6 \big ) \beta_3 \nonumber \\ & & \hspace*{38 mm } - 5 \big ( 945 - 210/q^2 - 136/q^4 - 110/q^6 - 105/q^8 \big )
\beta_4 \bigg ] , \\
\lambda_2 & = & \frac{\delta}{360 \mu } \bigg [ 1920 \beta_1 - 360 \big ( 5 - 1/q^2 \big ) \beta_2 + 48 \big ( 35 - 6/q^2 - 5/q^4 \big ) \beta_3 - 5 \big ( 315 - 49/q^2 - 39/q^4 - 35/q^6 \big ) \beta_4 \bigg ] , \\
\lambda_3 & = & \hspace*{0.9 mm } \frac{\delta}{90 \mu } \hspace*{0.9 mm } \bigg [ 480 \beta_1 -360 \beta_2 + 48 \big ( 7 - 1/q^2 \big ) \beta_3 - 5 \big ( 63 - 8/q^2 - 7/q^4 \big ) \beta_4 \bigg ] , \\ \lambda_4 & = & \hspace*{0.9 mm } \frac{\delta}{15 \mu } \hspace*{0.9 mm } \bigg [ -60 \beta_2 + 48 \beta_3 - 5 \big ( 9 - 1/q^2 \big ) \beta_4 \bigg ] , \\
\lambda_5 & = & \hspace*{0.9 mm } \frac{8 \delta}{15 \mu } \hspace*{0.9 mm } \bigg [ 6 \beta_3 - 5 \beta_4 \bigg ] , \\
\lambda_6 & = & - \frac{8 \delta}{3 \mu } \beta_4.\end{aligned}\ ] ] to use this general solution to investigate the ( anti)rigid nature of these helicoidal trajectories , we need to observe how the ricci curvature @xmath265 depends on the correction . for simplicity , we take @xmath29 , and note that for the background solution @xmath266 now , suppose we wish to investigate the behaviour of the curvature near a general initial point @xmath267 , where @xmath268 .
then near @xmath267 @xmath269 where we have used ( [ helical : fulleom ] ) to evaluate @xmath270 , and noted that @xmath271 for @xmath29 .
now , the combination @xmath272 will be negative if if the magnitude of the curvature is decreased , which corresponds to an intuitive notion of rigidity . from ( [ eq : appcurv1 ] ) and ( [ eq : appcurv2 ] ) we see that this requires : @xmath273
\\ \left ( - \delta^2 + 2 \delta^4 - 8 \delta^4 \sigma^2 + 13 \delta^6 \sigma^2 - 6 \delta^6 \sigma^4
\right ) > 0\ ] ] where @xmath274 .
figure [ fig : curv ] shows the sign of @xmath272 as a function of the two parameters @xmath275 and @xmath210 .
the shaded zones indicate the regions where the string is antirigid .
we see that with the exception of the loop case ( @xmath276)the string admits both rigid and antirigid behaviour for each value of @xmath210 .
e - mail : [email protected] e - mail : [email protected] e - mail : [email protected] e - mail : [email protected] a. vilenkin and e. p. s. shellard , _ cosmic strings and other topological defects _ ( cambridge university press , cambridge , 1994 ) .
t. vachaspati and a. achcarro , phys .
d * 44 * , 3067 ( 1991 ) ; m. hindmarsh , phys .
lett . * 68 * , 1263 ( 1991 ) . t. vachaspati , phys . rev . lett . *
68 * , 1977 ( 1992 ) ; phys . rev .
lett . * 69 * , 216(e ) ( 1992 ) ; n. manton , phys .
d * 28 * , 2019 ( 1983 ) .
n. turok , phys .
* 63 * , 2625 ( 1989 ) .
r. brandenberger , _ modern cosmology and structure formation _
[ astro - ph/9411049 ] ; m. hindmarsh and t. w. b. kibble , rep .
. phys . * 58 * , 477 ( 1995 ) [ hep - ph/9411342 ] .
s. shellard , nucl .
b * 283 * , 624 ( 1987 ) ; k. j. m. moriarty , e. myers and c. rebbi , phys
. lett . * 207 * , 411 ( 1988 ) .
a. vilenkin , phys .
d * 23 * , 852 ( 1981 ) ; j. r. gott iii , astrophys .
j. * 288 * , 422 ( 1985 ) ; w. hiscock , phys . rev .
d * 31 * , 3288 ( 1985 ) ; b. linet , gen . relativ
* 17 , * 1109 ( 1985 ) ; d. garfinkle , phys .
d * 32 * , 1323 ( 1985 ) ; r. gregory , phys .
* 59 , * 740 ( 1987 ) .
t. vachaspati and a. vilenkin , phys .
lett . * 67 * , 1057 ( 1991 ) ; d. n. vollick , phys .
d * 45 * , 1884 ( 1992 ) .
maeda and n. turok , phys .
* 202*b , 376 ( 1988 ) ; r. gregory , phys . lett . * 206*b , 199 ( 1988 ) .
d. garfinkle and r. gregory , phys .
d * 41 * , 1889 ( 1990 ) ; r. gregory , d. haws and d. garfinkle , phys .
d * 42 * , 343 ( 1990 ) ; r. gregory , phys .
d * 43 * , 520 ( 1991 ) . v.
silveira and m. d. maia , phys .
lett . * 174*a , 280 ( 1993 ) [ gr - qc/9303017 ] ; p. orland , nucl .
b * 428 * , 221 ( 1994 ) [ hep - th/9404140 ] ; h. arod , nucl .
b * 450 * , 174 ( 1995 ) [ hep - th/9502018 ] ; nucl . phys .
b * 450 * , 189 ( 1995 ) [ hep - th/9503001 ] . h. b. nielsen and p. olesen , nucl .
b * 61 * , 45 ( 1973 ) .
d. frster , nucl .
b * 81 * , 84 ( 1974 ) . b. carter and r. gregory , phys
d * 51 * , 5839 ( 1995 ) [ hep - th/9410095 ] .
m. anderson , phys .
d * 51 * , 2863 ( 1995 ) .
t. l. curtright , g. i. ghandour and c. k. zachos , phys .
d * 34 * , 3811 ( 1986 ) ; b. boisseau and p. s. letelier , phys .
d * 46 * , 1721 ( 1992 ) ; h. arod and a. l. larsen , phys .
d * 49 * , 4154 ( 1994 ) [ hep - th/9309089 ] .
a. polyakov , nucl .
b * 268 * , 406 ( 1986 ) .
h. kleinert , phys . lett . * 174*b , 335 ( 1986 ) . b. carter , class .
quantum grav . *
11 * , 2677 ( 1994 ) .
t. w. b. kibble and n. turok , phys . lett . * 116*b , 141 ( 1982 ) .
d. garfinkle and t. vachaspati , phys .
d * 42 * , 1960 ( 1990 ) .
m. sakellariadou , phys .
d * 42 * , 453 ( 1990 ) ; phys .
d * 43 * , 4150(e ) ( 1991 ) .
m. spivak , _ differential geometry _
( publish or perish , berkeley , ca , 1979 ) .
b. carter , j. geom . phys . * 8 * , 53 ( 1992 ) ; [ hep - th/9705172 ] .
numerical algorithms group , _ nag fortran library mark 16 _ ( oxford , 1993 ) . w. h. press , s. a. teukolski , w. t. vetterling and b. p. flannery , _ numerical recipes in c : the art of scientific computing _
( cambridge university press , cambridge , 1992 ) .
ddddd @xmath118 & @xmath277 & @xmath278 & @xmath279 & @xmath280 + 0.1 & 3.272 & 6.025 & 77.761 & -104.985 + 0.2 & 2.813 & 3.347 & 23.565 & -32.835 + 0.5 & 2.314 & 1.634 & 4.308 & -6.767 + 1.0 & 2.000 & 0.999 & 1.069 & -2.138 + 2.0 & 1.736 & 0.642 & 0.174 & -0.722 + 10.0 & 1.278 & 0.272 & -0.097 & -0.094 + 50.0 & 0.977 & 0.141 & -0.065 & -0.030 + 100.0 & 0.880 & 0.111 & -0.052 & -0.023 + | we examine the leading order corrections to the nambu effective action for the motion of a cosmic string , which appear at fourth order in the ratio of the width to radius of curvature of the string .
we determine the numerical coefficients of these extrinsic curvature corrections , and derive the equations of motion of the worldsheet . using these equations , we calculate the corrections to the motion of a collapsing loop , a travelling wave , and a helical breather . from the numerical coefficients we have calculated , we discuss whether the string motion can be labelled as ` rigid ' or ` antirigid , ' and hence whether cusp or kink formation might be suppressed or enhanced . |
solving large scale eigenvalue problems becomes a fundamental problem in modern science and engineering society .
however , it is always a very difficult task to solve high - dimensional eigenvalue problems which come from physical and chemistry sciences .
xu and zhou @xcite give a type of two - grid discretization method to improve the efficiency of the solution of eigenvalue problems .
by the two - grid method , the solution of eigenvalue problem on a fine mesh is reduced to a solution of eigenvalue problem on a coarse mesh ( depends on the fine mesh ) and a solution of the corresponding boundary value problem on the fine mesh @xcite . for more details , please read @xcite .
combing the two - grid idea and the local and parallel finite element technique @xcite , a type of local and parallel finite element technique to solve the eigenvalue problems is given in @xcite ( also see @xcite ) .
recently , a new type of multilevel correction method for solving eigenvalue problems which can be implemented on multilevel grids is proposed in @xcite . in the multilevel correction scheme
, the solution of eigenvalue problem on a finest mesh can be reduced to a series of solutions of the eigenvalue problem on a very coarse mesh ( independent of finest mesh ) and a series of solutions of the boundary value problems on the multilevel meshes .
the multilevel correction method gives a way to construct a type of multigrid scheme for the eigenvalue problem @xcite .
in this paper , we propose a type of multilevel local and parallel scheme to solve the eigenvalue problem based on the combination of the multilevel correction method and the local and parallel technique .
an special property of this scheme is that we can do the local and parallel computing for any times and then the mesh size of original coarse triangulation is independent of the finest triangulation . with this new method
, the solution of the eigenvalue problem will not be more difficult than the solution of the boundary value problems by the local and parallel algorithm since the main part of the computation in the multilevel local and parallel method is solving the boundary value problems .
the standard galerkin finite element method for eigenvalue problem has been extensively investigated , e.g. babuka and osborn @xcite , chatelin @xcite and references cited therein .
there also exists analysis for the local and parallel finite element method for the boundary value problems and eigenvalue problems @xcite .
here we adopt some basic results in these papers for our analysis . the corresponding error and computational work estimates of the proposed multilevel local and parallel scheme for the eigenvalue problem
will be analyzed .
based on the analysis , the new method can obtain optimal errors with an optimal computational work in each processor . an outline of this paper goes as follows .
in the next section , a basic theory about the local error estimate of the finite element method is introduced . in section 3
, we introduce the finite element method for the eigenvalue problem and the corresponding error estimates .
a local and parallel type of one correction step and multilevel correction algorithm will be given in section 4 .
the estimate of the computational work for the multilevel local and parallel algorithm is presented in section 5 . in section 6 ,
two numerical examples are presented to validate our theoretical analysis and some concluding remarks are given in the last section .
in this section , we introduce some notation and error estimates of the finite element approximation for linear elliptic problem .
the letter @xmath0 ( with or without subscripts ) denotes a generic positive constant which may be different at its different occurrences through the paper . for convenience
, the symbols @xmath1 , @xmath2 and @xmath3 will be used in this paper . that @xmath4 and @xmath5 , mean that @xmath6 , @xmath7 and @xmath8 for some constants @xmath9 and @xmath10 that are independent of mesh sizes ( see , e.g. , @xcite ) .
we shall use the standard notation for sobolev spaces @xmath11 and their associated norms and seminorms ( see , e.g. , @xcite ) . for @xmath12 ,
we denote @xmath13 and @xmath14 , where @xmath15 is in the sense of trace , @xmath16 . for @xmath17
, the notation @xmath18 means that @xmath19 ( see figure [ fig : domaindg ] ) .
it is well known that any @xmath20 can be naturally extended to be a function in @xmath21 with zero outside of @xmath22 , where @xmath23 .
thus we will show this fact by the abused notation @xmath24 .
, width=377,height=170 ] now , let us define the finite element space .
first we generate a shape - regular decomposition @xmath25 of the computing domain @xmath26 into triangles or rectangles for @xmath27 ( tetrahedrons or hexahedrons for @xmath28 ) .
the diameter of a cell @xmath29 is denoted by @xmath30 .
the mesh size function is denoted by @xmath31 whose value is the diameter @xmath30 of the element @xmath32 including @xmath33 . for generality , following @xcite , we shall consider a class of finite element spaces that satisfy certain assumptions .
now we describe such assumptions . * a.0*. there exists @xmath34 such that @xmath35 where @xmath36 is the largest mesh size of @xmath25 . based on the triangulation @xmath25 , we define the finite element space @xmath37 as follows @xmath38 where @xmath39 denotes the space of polynomials of degree not greater than a positive integer @xmath40 .
then we know @xmath41 and define @xmath42 .
given @xmath43 , we define @xmath44 and @xmath45 to be the restriction of @xmath37 and @xmath25 to @xmath46 , respectively , and @xmath47 for any @xmath43 mentioned in this paper , we assume that it aligns with the partition @xmath25 . as we know , the finite element space @xmath48 satisfy the following proposition ( see , e.g. , @xcite ) .
[ proposition_fractional_norm](_fractional norm _ ) for any @xmath43 , we have @xmath49 in this subsection , we repeat some basic properties of a second order elliptic boundary value problem and its finite element discretization , which will be used in this paper .
the following results is presented in @xcite .
we consider the homogeneous boundary value problem @xmath50 here the linear second order elliptic operator @xmath51 is define as @xmath52 where @xmath53 is uniformly positive definite symmetric on @xmath54 with @xmath55 .
the weak form for ( [ liner_elliptic ] ) is as follows : find @xmath56 such that @xmath57 where @xmath58 is the standard inner - product of @xmath59 and @xmath60 as we know @xmath61 we assume ( c.f .
@xcite ) that the following regularity estimate holds for the solution of ( [ liner_elliptic ] ) or ( [ weak_linear_elliptic ] ) @xmath62 for some @xmath63 $ ] depending on @xmath54 and the coefficient of @xmath64 . for some @xmath43
, we need the following regularity assumption * r(g)*. for any @xmath65 , there exists a @xmath66 satisfying @xmath67 and @xmath68 for the analysis , we define the galerkin - projection operator @xmath69 by @xmath70 and apparently @xmath71 based on ( [ projection_inequality ] ) , the global priori error estimate can be obtained from the approximate properties of the finite dimensional subspace @xmath72 ( cf .
@xcite ) . for the following analysis
, we introduce the following quantity : @xmath73 similarly , we can also define @xmath74 if assumption r(g ) holds .
the following results can be found in @xcite .
@xmath75 now , we state an important and useful result about the local error estimates @xcite which will be used in the following . [ prop : local_estimate ] suppose that @xmath76 and @xmath77 .
if assumptions a.0 holds and @xmath78 satisfies @xmath79 then we have the following estimate @xmath80
in this section , we introduce the concerned eigenvalue problem and the corresponding finite element discretization . in this paper , we consider the following eigenvalue problem : find @xmath81 such that @xmath82 and @xmath83 where @xmath84 for the eigenvalue @xmath85 , there exists the following rayleigh quotient expression ( see , e.g. , @xcite ) @xmath86 from @xcite , we know the eigenvalue problem ( [ weak_eigenvalue_problem ] ) has an eigenvalue sequence @xmath87 @xmath88 and the associated eigenfunctions @xmath89 where @xmath90 . in the sequence @xmath91 , the @xmath92 are repeated according to their multiplicity
. then we can define the discrete approximation for the exact eigenpair @xmath93 of ( [ weak_eigenvalue_problem ] ) based on the finite element space as : find @xmath94 such that @xmath95 and @xmath96 from ( [ discrete_weak_eigen_problem ] ) , we know the following rayleigh quotient expression for @xmath97 holds ( see , e.g. , @xcite ) @xmath98 similarly , we know from @xcite the eigenvalue problem ( [ discrete_weak_eigen_problem ] ) has eigenvalues @xmath99 and the corresponding eigenfunctions @xmath100 where @xmath101 ( @xmath102 is the dimension of the finite element space @xmath72 ) . from the minimum - maximum principle ( see , e.g. , @xcite ) , the following upper bound result holds @xmath103 let @xmath104 denote the eigenspace corresponding to the eigenvalue @xmath105 which is defined by @xmath106 where @xmath107
. then we define @xmath108 for the eigenpair approximations by the finite element method , there exist the following error estimates . ( ( * ? ? ?
* lemma 3.7 , ( 3.28b,3.29b ) ) , ( * ? ? ?
* p. 699 ) and @xcite)[prop : eigen_error_estimate ]
\(i ) for any eigenfunction approximation @xmath109 of ( [ discrete_weak_eigen_problem ] ) @xmath110 , there is an eigenfunction @xmath111 of ( [ weak_eigenvalue_problem ] ) corresponding to @xmath105 such that @xmath112 and @xmath113 furthermore , @xmath114 ( ii ) for each eigenvalue , we have @xmath115 here and hereafter @xmath116 is some constant depending on @xmath117 but independent of the mesh size @xmath118 . to analyze our method
, we introduce the error expansion of eigenvalue by the rayleigh quotient formula which comes from @xcite .
[ prop : rayleigh_quotient_error ] assume @xmath93 is the true solution of the eigenvalue problem ( [ weak_eigenvalue_problem ] ) and @xmath119 .
let us define @xmath120 then we have @xmath121
in this section , we present a new multilevel parallel algorithm to solve the eigenvalue problem based on the combination of the local and parallel finite element technique and the multilevel correction method .
first , we introduce an one correction step with the local and parallel finite element scheme and then present a parallel multilevel method for the eigevalue problem . for the description of the numerical scheme , we need to define some notation . given an coarsest triangulation @xmath122
, we first divide the domain @xmath54 into a number of disjoint subdomains @xmath123 , @xmath124 , @xmath125 such that @xmath126 , @xmath127 ( see figure [ fig : domaingi ] ) , then enlarge each @xmath128 to obtain @xmath129 that aligns with @xmath122 .
we pick another sequence of subdomains @xmath130 and ( see figure [ fig : domaingi ] ) @xmath131 , width=377,height=170 ] in this paper we assume the domain decomposition satisfies the following property @xmath132 for any @xmath133 with @xmath134 .
first , we present the one correction step to improve the accuracy of the given eigenvalue and eigenfunction approximation .
this correction method contains solving an auxiliary boundary value problem in the finer finite element space on each subdomain and an eigenvalue problem on the coarsest finite element space .
for simplicity of notation , we set @xmath135 and @xmath136 to denote an eigenpair and its corresponding approximation of problems ( [ weak_eigenvalue_problem ] ) and ( [ discrete_weak_eigen_problem ] ) , respectively . for the clear understanding
, we only describe the algorithm for the simple eigenvalue case . the corresponding algorithm for the multiple eigenvalue case
can be given in the similar way as in @xcite . in order to do the correction step ,
we build original coarsest finite element space @xmath137 on the background mesh @xmath122 .
this coarsest finite element space @xmath137 will be used as the background space in our algorithm .
assume we have obtained an eigenpair approximation @xmath138 .
the one correction step will improve the accuracy of the current eigenpair approximation @xmath139 .
let @xmath140 ba a finer finite element space such that @xmath141 .
here we assume the finite element spaces @xmath142 and @xmath140 are consistent with the domain decomposition and @xmath143 .
based on this finer finite element space @xmath140 , we define the following one correction step .
[ algm : one_step_correction ] one correction step we have a given eigenpair approximation @xmath138 . 1 .
define the following auxiliary boundary value problem : + for each @xmath144 , find @xmath145 such that @xmath146 set @xmath147 .
2 . construct @xmath148 such that @xmath149 in @xmath150 @xmath151 and @xmath152 in @xmath153 with @xmath154 being defined by solving the following problem : + find @xmath155 such that @xmath156 @xmath151 and @xmath157 3 .
define a new finite element space @xmath158 and solve the following eigenvalue problem : + find @xmath159 such that @xmath160 and @xmath161 summarize the above three steps into @xmath162 where @xmath163 and @xmath164 are the given eigenvalue and eigenfunction approximation , respectively .
[ thm : error_estimate_one_step_correction ] assume the current eigenpair approximation @xmath138 has the following error estimates @xmath165 has the following error estimates @xmath166 .
we focus on estimating @xmath167 .
first , we have @xmath168 and @xmath169 from problems ( [ projection_problem ] ) , ( [ weak_eigenvalue_problem ] ) and ( [ aux_problem ] ) , the following equation holds @xmath170 for @xmath171 . according to proposition [ prop : local_estimate ]
@xmath172 we will estimate the first term , i.e. @xmath173 by using the aubin - nitsche duality argument . given any @xmath174 ,
there exists @xmath175 such that @xmath176 let @xmath177 and @xmath178 satisfying @xmath179 then the following equations hold @xmath180 where @xmath143 and ( [ projection_problem ] ) , ( [ weak_eigenvalue_problem ] ) , ( [ discrete_weak_eigen_problem ] ) , ( [ aux_problem ] ) are used in the last equation .
combining ( [ equation_nitsche ] ) and the following error estimates @xmath181 we have @xmath182 from ( [ equ : key2 ] ) and ( [ equ : key5 ] ) , for @xmath183 , we have @xmath184 now , we estimate @xmath185 . from ( [ projection_problem ] ) , ( [ weak_eigenvalue_problem ] ) and ( [ g_m+1_problem ] ) , we obtain @xmath186 for any @xmath187 , the following estimates hold @xmath188 combining ( [ estimate_g_m+1 ] ) and the following estimate @xmath189 and proposition [ proposition_fractional_norm ] , we have @xmath190 combining ( [ domain_decomposition_property ] ) , ( [ equ : key1 ] ) , ( [ equ : key3 ] ) and ( [ equ : key4 ] ) leads to @xmath191 together with the error estimate of the finite element projection @xmath192 and ( [ estimate_lambda_lambda_h_k ] ) , ( [ equ : key0 ] ) , we have @xmath193 from ( [ error_u_h_k+1 ] ) and ( [ error_tilde_u_h_k+1 ] ) , we can obtain ( [ estimate_u_u_h_k+1 ] ) . we come to estimate the error for the eigenpair solution @xmath194 of problem ( [ eigen_augment_problem ] ) . based on the error estimate theory of eigenvalue problems by finite element methods ( see , e.g. ,
proposition [ prop : eigen_error_estimate ] or ( * ? ? ?
* theorem 9.1 ) ) and the definition of the space @xmath195 , the following estimates hold @xmath196 and @xmath197 where @xmath198 so we obtain the desired result ( [ estimate_u_u_h_k+1 ] ) , ( [ estimate_u_u_h_k+1_zero ] ) and the estimate ( [ estimate_lambda_lambda_h_k+1 ] ) can be obtained by proposition [ prop : rayleigh_quotient_error ] and ( [ estimate_u_u_h_k+1 ] ) .
now we introduce a type of multilevel local and parallel scheme based on the one correction step defined in algorithm [ algm : one_step_correction ] .
this type of multilevel method can obtain the same optimal error estimate as solving the eigenvalue problem directly in the finest finite element space . in order to do multilevel local and parallel scheme ,
we define a sequence of triangulations @xmath199 of @xmath54 determined as follows .
suppose @xmath200 is obtained from @xmath122 by the regular refinement and let @xmath199 be obtained from @xmath201 via regular refinement ( produce @xmath202 congruent elements ) such that @xmath203 based on this sequence of meshes , we construct the corresponding linear finite element spaces such that for each @xmath144 @xmath204 and the following relation of approximation errors holds @xmath205 the relation ( [ error_k_k_1 ] ) is reasonable since we can choose @xmath206 .
always the upper bound of the estimate @xmath207 holds .
recently , we also obtain the lower bound @xmath208 ( c.f .
@xcite ) .
[ algm : multi_correction ] multilevel correction scheme 1 .
solve the following eigenvalue problem in @xmath209 : + find @xmath210 such that @xmath211 and @xmath212 2 .
construct a series of finer finite element spaces @xmath213 such that @xmath214 and ( [ error_k_k_1 ] ) holds .
3 . do @xmath215 * obtain a new eigenpair approximation @xmath216 by algorithm [ algm : one_step_correction ] @xmath217 + end do finally , we obtain an eigenpair approximation @xmath218 .
[ thm : multi_correction ] after implementing algorithm [ algm : multi_correction ] , there exists an eigenfunction @xmath219 such that the resultant eigenpair approximation @xmath220 has the following error estimate @xmath221 for some constant @xmath0 .
based on proposition [ prop : eigen_error_estimate ] , there exists an eigenfunction @xmath219 such that @xmath222 let @xmath223 . from ( [ initial_error_eigenvalue])-([initial_error_eigenfunc_0 ] ) and theorem [ thm : error_estimate_one_step_correction ]
, we have @xmath224 by a process of induction with the condition @xmath225 .
then by recursive relation , we obtain @xmath226 based on the proof in theorem [ thm : error_estimate_one_step_correction ] , ( [ error_k_k_1 ] ) and ( [ varepsilon_h_n ] ) , the final eigenfunction approximation @xmath227 has the error estimate @xmath228 the desired result ( [ multi_correction_err_fun0 ] ) and ( [ multi_correction_err_eigen ] ) can also be proved with the similar way in the proof of theorem [ thm : error_estimate_one_step_correction ] .
in this section , we turn our attention to the estimate of computational work for algorithm [ algm : multi_correction ] . we will show that algorithm [ algm : multi_correction ] makes solving eigenvalue problem need almost the same work as solving the boundary value problem by the local and parallel finite element method .
first , we define the dimension of each level linear finite element space as @xmath229 then we have @xmath230 assume the eigenvalue problem solving in the coarsest spaces @xmath137 and @xmath209 need work @xmath231 and @xmath232 , respectively , and the work of solving the boundary value problem in @xmath233 and @xmath234 be @xmath235 and @xmath236 , @xmath237 .
then the work involved in algorithm [ algm : multi_correction ] is @xmath238 for each processor .
furthermore , the complexity in each processor will be @xmath239 provided @xmath240 and @xmath241 .
let @xmath242 denote the work in any processor of the one correction step in the @xmath40-th finite element space @xmath243 .
then with the definition , we have @xmath244 iterating ( [ work_k ] ) and using the fact ( [ relation_dimension ] ) , we obtain @xmath245 this is the desired result @xmath238 and the one @xmath239 can be obtained by the conditions @xmath240 and @xmath241 .
the linear complexity @xmath235 and @xmath236 can be arrived by the so - called multigrid method ( see , e.g. , @xcite ) .
in this section , we give two numerical examples to illustrate the efficiency of the multilevel correction algorithm ( algorithm [ algm : multi_correction ] ) proposed in this paper .
[ example_1 ] in this example , the eigenvalue problem ( [ weak_eigenvalue_problem ] ) is solved on the square @xmath246 with @xmath247 and @xmath248 . as in figure
[ fig : computedomain ] , we first divide the domain @xmath54 into four disjoint subdomains @xmath123 , @xmath124 , @xmath249 such that @xmath250 , @xmath127 , then enlarge each @xmath128 to obtain @xmath129 such that @xmath251 for @xmath252 and @xmath253 , @xmath254 , width=302,height=151 ] the sequence of finite element spaces is constructed by using the linear or quadratic element on the nested sequence of triangulations which are produced by the regular refinement with @xmath255 ( connecting the midpoints of each edge ) .
algorithm [ algm : multi_correction ] is applied to solve the eigenvalue problem .
if the linear element is used , from theorem [ thm : multi_correction ] , we have the following error estimates for eigenpair approximation @xmath256-st , @xmath257-th eigenfunctions ( they are simple ) by the linear finite element method with five levels grids are shown in tables [ tab : num_res_eva ] and [ tab : num_res_eve ] . it is observed from tables [ tab : num_res_eva ] and [ tab : num_res_eve ] that the numerical results confirm the efficiency of the proposed algorithm .
.the errors for the first @xmath258 eigenvalue approximations [ cols="^,^,^,^,^,^",options="header " , ] next we discuss the effectiveness of @xmath259 and the coarsest mesh size @xmath260 to the numerical results by algorithm [ algm : multi_correction ] . figure [ fig : p1hd ] shows the errors for the different choices of @xmath259 and @xmath260 by the linear finite element method . from figure
[ fig : p1hd ] , we can find algorithm [ algm : multi_correction ] can obtain the optimal convergence order when @xmath261 and @xmath262 which are soft requirements .
eigenvalue approximations by the linear element : the left subfigure is for @xmath263 and @xmath264 , @xmath265 , @xmath266 .
the right subfigure is for @xmath267 and @xmath263 , @xmath268 , @xmath269,title="fig:",width=188,height=188 ] eigenvalue approximations by the linear element : the left subfigure is for @xmath263 and @xmath264 , @xmath265 , @xmath266 .
the right subfigure is for @xmath267 and @xmath263 , @xmath268 , @xmath269,title="fig:",width=188,height=188 ] the case becomes better when we use the quadratic finite element method ( see figure [ fig : p2hd ] ) . for the quadratic finite element ,
the convergence order ( @xmath257-th ) is always optimal even when @xmath259 is very small .
eigenvalue approximations by the quadratic element : the left subfigure is for @xmath270 and @xmath267 , @xmath265 , @xmath266 .
the right subfigure is for @xmath267 and @xmath271 , @xmath269 , @xmath272,title="fig:",width=188,height=188 ] eigenvalue approximations by the quadratic element : the left subfigure is for @xmath270 and @xmath267 , @xmath265 , @xmath266 .
the right subfigure is for @xmath267 and @xmath271 , @xmath269 , @xmath272,title="fig:",width=188,height=188 ] [ example_2 ] in the second example , we solve the eigenvalue problem ( [ weak_eigenvalue_problem ] ) using linear and quadratic element on the square @xmath246 with @xmath273 , @xmath274 and @xmath275 since the exact eigenvalue is not known , we use the accurate enough approximations @xmath276 $ ] by the extrapolation method as the first @xmath277 exact eigenvalues to investigate the errors . eigenvalue approximations with @xmath278 and @xmath279 : the left subfigure is for linear element and the right subfigure is for quadratic element , title="fig:",width=188,height=188 ] eigenvalue approximations with @xmath278 and @xmath279 : the left subfigure is for linear element and the right subfigure is for quadratic element , title="fig:",width=188,height=188 ] figure [ fig : p1p2 ] shows the corresponding numerical results for the first @xmath277 eigenvalues by the linear and quadratic finite element methods , respectively . here , we use four level grids to do the numerical experiments . from figure
[ fig : p1p2 ] , the numerical results also confirm the efficiency of the proposed algorithm in this paper .
in this paper , we give a new type of multilevel local and parallel method based on multigrid discretization to solve the eigenvalue problems .
the idea here is to use the multilevel correction method to transform the solution of eigenvalue problem to a series of solutions of the corresponding boundary value problems with the local and parallel method .
as stated in the numerical examples , algorithm [ algm : multi_correction ] for simple eigenvalue cases can be extended to the corresponding version for multiple eigenvalue cases . for more information
, please refer @xcite .
furthermore , the framework here can also be coupled with the adaptive refinement technique .
the ideas can be extended to other types of linear and nonlinear eigenvalue problems .
these will be investigated in our future work .
i. babuka and j. osborn , _ eigenvalue problems _ , in handbook of numerical analysis , vol .
ii , ( eds .
p. g. lions and ciarlet p.g . ) , finite element methods ( part 1 ) , north - holland , amsterdam , 641 - 787 , 1991 . | a local and parallel algorithm based on the multilevel discretization is proposed in this paper to solve the eigenvalue problem by the finite element method . with this new scheme , solving the eigenvalue problem in the finest grid is transferred to solutions of the eigenvalue problems on the coarsest mesh and a series of solutions of boundary value problems by using the local and parallel algorithm .
the computational work in each processor can reach the optimal order .
therefore , this type of multilevel local and parallel method improves the overall efficiency of solving the eigenvalue problem .
some numerical experiments are presented to validate the efficiency of the new method . 0.3 cm * keywords . *
eigenvalue problem , multigrid , multilevel correction , local and parallel method , finite element method .
0.2 cm * ams subject classifications .
* 65n30 , 65n25 , 65l15 , 65b99 . |
the possibility that supermassive black holes ( sbhs ) inhabit the centers of many if not most galaxies , and the observed correlation between sbh masses and galactic bulge properties , has potentially a fundamental significance for our understanding of galaxy formation and evolution .
the relationships between black hole and bulge properties include a loose relationship between sbh and bulge masses , @xmath5 , and an apparently much tighter one between the sbh mass and the velocity dispersion in the corresponding bulge , @xmath6 ( e.g. , ferrarese & merritt 2000 ; gebhardt et al . 2000 ; tremaine et al .
2002 ; cf .
reviews by kormendy & gebhardt 2001 ; merritt & ferrarese 2001 ) . in this paper we attempt to provide a physical explanation for these relationships between sbhs and their host galaxies .
our model is based on the interaction between the dark haloes of galaxies and the baryonic components settling in their midst . as baryonic matter accumulates to form the bulge and sbh , the orbital structure of the underlying gravitational potential is modified .
this , in turn , affects the subsequent accumulation of gas , which is highly dissipative and therefore is sensitive to orbital geometry .
consequently , halo properties will determine those of the bulge and the sbh , and this will give rise to correlations between them that are compatible with the observed ones .
cuspy , triaxial haloes appear to be a natural outcome of dissipationless collapse in a cold dark matter ( cdm ) dominated universe ( e.g. , cole & lacey 1996 ) .
interactions with the baryonic component during the initial stages of collapse may affect the triaxiality , making it milder but still non - negligible ( dubinsky 1994 ) .
before it even becomes dominant in the central regions , a clumpy baryonic component can also level off the central cusps of dark haloes , producing harmonic cores ( el - zant , shlosman & hoffman 2001 ) .
even though this may affect the equidensity contours , again making them rounder , it need not symmetrize the equipotentials ; these can remain asymmetric if triaxiality is not affected beyond some radius ( for example , a homogeneous bar has a non - axisymmetric force contribution inside its density figure ) . from an observational standpoint it appears that haloes of fully formed galaxies tend to have nearly constant density cores ( de blok & bosma 2002 ) and that residual potential axial ratios of about 0.9 in cdm haloes are plausible , _ even in present day galaxies _ ( e.g. , kuijken & tremaine 1994 ; rix & zaritsky 1995 ; rix 1995 ) .
the orbital structure of the inner regions of slowly rotating , non - axisymmetric potentials with harmonic cores is dominated by box orbits ( e.g. , binney & tremaine 1987 ) , which have no particular sense of circulation .
they are self - intersecting and , therefore , can not be populated with gas .
dissipation causes material to sink quickly toward the only long - lived attractor available the center ( e.g. , pfenniger & norman 1990 ; el - zant 1999 ) . in the process ,
interaction with the triaxial harmonic core causes the baryonic material to lose most of its angular momentum . the combined system would , therefore , also be slowly rotating . thus ,
unless star formation terminates the collapse , the final concentration of the first baryonic material could be extremely large .
the onset of star formation is expected to occur when the baryonic material becomes self - gravitating roughly speaking , when its density becomes larger than that of the halo core .
this also happens to be the criterion for the destruction of the harmonic core and the emergence of loop ( or tube ) orbits , which do have a definite sense of circulation .
the role of the sbh is to contribute to the emergence of these orbits in the very central region .
it is the collusion between the sbh and the more extended hot baryonic component in creating the loop orbits that leads to the correlations claimed in this paper . whereas , in three dimensions , a box orbit can be represented as a superposition of three ( generally ) incommensurable radial oscillations along mutually perpendicular axes , which are efficiently attenuated by dissipation , loop orbits are best described in terms of modest radial and vertical excursions superposed on rotational motion about the center . while the vertical and radial excursions , like the oscillations characterizing box orbits , are attenuated by dissipation ,
the rotational motion is not efficiently dissipated among gas clouds populating such orbits ( in the same sense of circulation ) .
the minimization of the radial and vertical oscillations results in closed periodic orbits that are confined to a plane .
we expect the amount of gas dissipation on loop orbits to depend on their axial ratios for , again , motion on highly eccentric orbits would lead to shocks and the accompanying loss of angular momentum on a short dynamical timescale , as observed in numerical simulations of gas flows in barred galaxies ( e.g. , heller & shlosman 1994 ) .
lacking a detailed model for the dissipation rate associated with gaseous motion , we will assume that there is a critical eccentricity _ below _ which the loop orbits can serve as long - lived attractors for dissipative motion .
in other words , gas populating these orbits will evolve only secularly and not dynamically .
fortunately , this dependence on critical eccentricity will turn out to be weak . in
the following we explore , within the above framework , the formation of galactic bulges and central compact objects .
we show that it is possible to deduce a well - defined linear correlation between the masses of the central sbhs and corresponding bulges in given dark matter haloes , with mass ratios comparable to those observed .
we demonstrate that , if dark matter halo cores follow a faber - jackson type relation ( faber & jackson 1976 ) between their masses and velocity dispersions , then a similar relation also applies to the bulges .
an @xmath4 relation between the sbh mass and the bulge velocity dispersion , which under certain assumptions can have smaller scatter than the @xmath7 relation , also follows naturally . within this framework ,
it is possible to make a number of testable predictions concerning the related structures of bulges , sbhs and their host haloes .
we discuss these in the final section .
formal aspects of the perturbation analysis applied to bulge - halo systems have been deferred to appendices a and b. preliminary results of this work have been reported by shlosman ( 2002 ) .
since , at this stage , we are interested in _ generic _ dynamical phenomena related to orbital structure , the exact form of the halo potential is immaterial as long as it exhibits a harmonic core where no loop orbits can exist .
the exact distribution of the baryonic material is also not crucial , except , again , for its central density distribution .
if this diverges , say as @xmath8 , loop orbits will be created all the way to the center , since the bulge - halo system no longer possesses a harmonic core . in this case
there is no need to form the sbh .
if , on the other hand , the ( proto)-bulge has a harmonic core of its own , there will still be a nearly constant density region near the center with only box orbits unless a central point mass is present .
we will assume that ` cuspy ' bulges , when they exist , are products of later evolution ( for example , a result of star formation and cold dissipationless collapse : e.g. , lokas & hoffman 2000 ) .
we use a logarithmic standard form ( e.g. , binney & tremaine 1987 ) to represent the halo potential : @xmath9 where @xmath10 is the asymptotic ( in the limit @xmath11 and @xmath12 ) circular velocity , @xmath13 is the core radius and @xmath14 , @xmath15 are the potential axis ratios
. we will consider the process of bulge formation to be terminated , or at least substantially slowed down , when the combined ( baryonic plus halo ) potential admits sufficiently round non - intersecting loop orbits , which permit the long - term circulation of gas without excessive dissipation .
such motions necessarily take place in a symmetry plane determined by the ( orbit - averaged ) angular momentum ( e.g. , frank , king & raine 2002 ) .
therefore , for our purposes , it will suffice to consider only orbits in the plane @xmath16 and to ignore the vertical dimension .
we fix @xmath14 at @xmath17 ( though the effects of its variation are discussed where relevant ) and adopt a threshold axial ratio @xmath18 for orbits that can be populated with gas .
the value of @xmath18 that best describes when this happens remains to be investigated , but its mere existence is what is important here . as we show below , our results are rather insensitive to the exact value of @xmath18 .
the halo core mass will be taken to be @xmath19 and we define the halo density to be @xmath20 .
this is conveniently close to the value of the density in the region where the potential is effectively harmonic that is , within the region where , in the absence of the bulge component , no loop orbits exist for @xmath21 .
we note , however , that a bulge with a larger core radius probes a region of the halo core with smaller average density than the region probed by a bulge with a relatively small core radius .
this effect will be discussed in section 4 .
since our aforementioned criterion depends on the potential in a chosen symmetry plane , the exact three - dimensional mass distribution of the bulge is unimportant , as the same planar potential may arise from a variety of these .
a particularly convenient form for the potential in this situation is that of miyamoto & nagai ( 1975 ) , @xmath22 this potential can approximate a disk - bulge system , with the parameters @xmath23 and @xmath24 determining the scale length and height , respectively . in the symmetry plane , therefore , the potential can arise from a range of density distributions from highly flattened to spherical ones .
in general , when we refer to the `` bulge '' we will have a spherical system in mind with @xmath25 .
in particular , all densities quoted are calculated under this assumption .
we will comment on the effects of assuming a flattened distribution where relevant . in that case @xmath26 in eq .
( [ miyamoto ] ) is replaced by @xmath2 , the spherical bulge mass .
the bulge _ core _ is defined by the radius @xmath27 and encloses the core mass @xmath28 .
the bulge core density is defined as @xmath29 .
the scaling relations that arise do not depend on the absolute masses and lengths , but instead , on the ratios of these quantities among the various contributions to the potential .
an important property of the dynamics in our model is that _ any transformation that leaves the relative masses and lengths of the components constant will also not affect the orbital structure once scale transformations are taken into account .
_ such transformations only change the time units , hence the dynamics remains invariant on a new characteristic timescale .
in addition , _ in regions where the baryonic and halo components have the same mass distributions , i.e. , inside their cores , only the density ratios of these components determine their relative contributions to the potential , _ and hence the orbital structure . for the remainder of this paper
, we will describe all components of the potential in terms of units scaled to the dimensions of the halo potential , eq .
( [ logarithm ] ) .
we scale all distances to @xmath13 , with @xmath30 , and all masses to @xmath31 so that @xmath32 .
the normalized bulge core density is then @xmath33 .
the bulge potential in scaled units is @xmath34 the black hole potential , in the same units , is @xmath35 where @xmath36 .
note that all orbits in the total potential @xmath37 depend only on three parameters , which may be taken to be @xmath38 , @xmath39 , and @xmath40 .
in this section , we demonstrate that the arguments outlined above imply a linear relation between the sbh and bulge masses , provided that the baryonic and cdm density distributions are not completely axisymmetric and exhibit , at some stage , nearly constant density cores . fig .
[ homofin ] ( the result of orbital integrations ) exhibits the axis ratios of the closed loop orbits @xmath41 as a function of the longer axis length ( @xmath42 ) . in this figure , we have fixed the scaled bulge core density , @xmath43 , and the sbh - to - bulge mass ratio , @xmath44 , but vary the scaled bulge core radius , @xmath38 .
consequently , the bulge masses @xmath45 range over a factor of @xmath46 . as expected ,
when only the halo contributes to the potential ( dotted line ) , @xmath47 quickly drops to zero inside the halo core , since no loop orbits can exist there .
( for halo potential axis ratios smaller than the adopted value of 0.9 , i.e. , greater triaxiality , @xmath47 reaches 0 at radii that are progressively closer to the halo core radius , @xmath48 . as the baryonic component , both in the bulge and the sbh , becomes progressively more massive ,
correspondingly rounder loop orbits appear .
loop orbits in the very central region are produced by the presence of the sbh .
the latter is represented numerically by a plummer sphere with a softening scale of @xmath49 , in scaled units . near the center , where the sbh contribution is dominant ,
the axis ratios of the loop orbits tend to @xmath50 .
to2.5 in ' '' '' the axis ratio curves exhibit two distinct minima .
it is apparent that the inner minima of the curves in fig .
[ homofin ] have ( nearly ) the same values of @xmath47 , but occur at different radii @xmath51 .
this is a consequence of assuming a fixed ratio of sbh - to - bulge mass , @xmath40 , and a constant bulge core density @xmath52 .
because the bulge and halo densities are nearly uniform in the inner regions , their relative contributions to the potential are determined by their density ratio . in the absence of the sbh contribution , @xmath47 in fig .
[ homofin ] would tend to zero within the effective harmonic core of the bulge - halo system ( see eq .
[ [ expo ] ] ) .
when the relative density is kept constant ( and @xmath53 ) this is proportional to @xmath54 ( the effect of density variation will be examined in section [ inmin ] ) .
the reason for this is that inside the nearly constant - density bulge core , the potential , which is now a superposition of two nearly harmonic potentials , is nearly harmonic ( even if the new core is less triaxial than the halo alone , e.g. , if the bulge is assumed to be spherical ) .
the inner minima of the curves in fig .
[ homofin ] , therefore , correspond to a transition from the region where the sbh provides the dominant contribution toward the creation of loop orbits to that where the bulge provides this contribution .
therefore , the minima occur at radii where the gravitational acceleration due to the sbh is proportional to that due to the bulge+halo , @xmath55 ( since the minimum is located well within the bulge core radius ) .
since @xmath40 is taken to be constant , we expect the minima to occur at radii @xmath51 . in section 2
, we have defined the critical value @xmath18 above which gas circulation can be sustained for secular timescales .
here we have shown that given a critical value , @xmath18 , for the inner minimum and the relative bulge - to - halo core density ratio , a value of @xmath40 associated with this minimum is determined . however , satisfying the condition @xmath56 at the inner minimum does not imply that the same condition is satisfied at all radii within @xmath13 . until it is satisfied at the outer minimum , as well
, further evolution can occur .
this is briefly sketched below and outlined in more detail in section 3.4 . until the critical value of @xmath47 is reached at all radii inside the halo core
, the bulge will continue to grow , since outside the bulge core @xmath47 declines again .
it is clear from fig . 1
that , in contrast to that of the inner minimum , the value of @xmath47 at the outer minimum depends more sensitively on the bulge mass than on its density .
the growth of the bulge , therefore , determines a _ minimal _ bulge mass fixed by the condition that @xmath57 at all radii within the halo core .
if the sbh did not continue to grow along with the bulge , this growth in the bulge mass would decrease the linear correlation coefficient @xmath40 , but , as we show in section 3.4 , this leads only to near - independence of the final value of @xmath40 on @xmath18 and still keeps the values of @xmath40 within the observed range . as shown in the previous section
, the creation of loop orbits with axis ratio above a given value depends , in the central region , solely on the ratio of the bulge - to - halo core density and the mass ratio @xmath58 .
the actual value of the bulge mass determines only the position of the minimum ( as a fraction of the halo core radius ) , not the axis ratio at the minimum .
but even the radius of the minimum is largely insensitive to the bulge mass , i.e. , @xmath59 , provided that the minimum actually exists and the bulge core density relative to the halo harmonic core density remains constant .
we will now examine the bounds on @xmath0 and @xmath40 which ensure the existence of a minimum @xmath57 inside the bulge core and observe the effect of varying the bulge density .
we calculate and plot @xmath60 for different choices of @xmath18 ( fig .
[ compac ] ) , in the following way .
first , without any loss of generality , we choose a bulge mass @xmath45 that is large enough that the outer minimum satisfies @xmath61 , for all @xmath62 .
this is done in order to focus on the effects of the inner minima only .
( recall , from the previous section , that the value of the outer minimum depends on the mass of the bulge . ) fig .
[ compac ] , therefore , exhibits the effect of @xmath63 and @xmath40 on the inner minima of the @xmath47 curves at @xmath64 . in order to obtain @xmath60 , we vary @xmath0 , which is some fraction @xmath40 of the chosen bulge mass .
the density is varied by contracting the bulge ( by decreasing its core radius ) until the condition @xmath61 is satisfied everywhere inside the bulge core .
to2.5 in ' '' '' one observes in fig . 2 that at smaller densities @xmath40 tends toward an asymptotic ( and maximal ) value associated with a given @xmath18 .
we refer to this hereafter as the `` asymptotic '' regime . in the transition to this regime , along a @xmath65 curve
, the value of @xmath40 is monotonically increasing .
in other words , the bulge contribution toward the creation of loop orbits of a given eccentricity gradually diminishes , and is compensated by a greater contribution from the sbh component . in the process
, the minimum in @xmath47 moves outward .
the asymptotic regime corresponds to a situation in which the existence of loop orbits with the required elongation depends on the value of @xmath40 , irrespective of the bulge density ( that is , effectively , only on the sbh mass ) . in this limit
the sbh contribution to the potential is sufficient to create loop orbits with @xmath61 at all radii within the halo core ( i.e. , at @xmath66 ) , without additional contributions from the bulge component ( cf .
3 ) . essentially
, it corresponds to a rapid collapse to the center and formation of the sbh by a large fraction of the baryonic material , bypassing the formation of a bulge .
this regime appears to be of academic interest only , as it implies @xmath67 all the way to the center ; it will not be discussed further .
to2.5 in ' '' '' a second regime characterizes the dynamical state of the sbh - bulge - dark halo system with @xmath68 . in this
`` scaling regime '' the @xmath69 curves are parallel straight lines with average slopes of about @xmath70 .
this can be explained in the following manner . from eq .
( [ expo ] ) we know that the radius of the effective harmonic core of the bulge - halo system , @xmath42 , is proportional to @xmath71 . in the scaling regime
, the location of the minimum in @xmath47 will be proportional to @xmath42 .
now , in order to maintain the minimum at a specified value of @xmath18 , the gravitational acceleration due to the sbh must be proportional to the acceleration due to the halo ( which contains the nonaxisymmetry ) at @xmath42 .
this implies @xmath72 . substituting for @xmath42 and noting that @xmath73
, we obtain @xmath74 .
this is approximately what is found from figure 2 and it also holds if we change the mass of the bulge keeping the radius constant and again using eq .
( [ expo ] ) .
note , furthermore , that for constant @xmath39 this relation predicts the radius of the effective harmonic core ( and the location of the minimum ) to be proportional to @xmath75 , as expected from the heuristic considerations of the previous section .
while the asymptotic values of @xmath40 depend on @xmath2 , the values of @xmath40 in scaling regime do not .
the transition between these two regimes can , therefore , be characterized by a sharp change in the behavior of @xmath40 , as seen from fig . 2 .
we are mainly interested in the scaling region , because the onset of star formation can be tied to the baryonic component becoming self - gravitating , which corresponds roughly to the bulge core density exceeding that of the background halo , i.e. , @xmath68 . in this case , and for @xmath76 , the sbh contributes significantly to the potential only at radii @xmath77 .
the @xmath47 curves , therefore , exhibit a definite inner minimum well inside the halo core . the values of @xmath78 and the initial @xmath40 , in principle , also depend on the critical eccentricity , @xmath18 , of the inner minimum , which is expected to be independent of the bulge mass , but which does depend on complex gas dynamics . if , after the sbh forms , the bulge mass falls short of the value required for creating sufficently round closed loop orbits at all radii inside the halo core , it will continue to grow until the outer minimum of the axis ratio curve also attains @xmath61 . in section 3.4
, we show that the sbh growth , if continued , becomes intermittent . in the next section ,
we also demonstrate that one can place constraints on the possible range of @xmath40 values by considering this constraint on the outer minimum . to obtain loop orbits rounder than a given @xmath18 _
at all radii _
@xmath79 , and not only in the central region , one in fact needs to take into account the bulge - to - halo core mass ratio and not only the density ratio .
this can already be seen in fig .
[ homofin ] , where all curves with sbh and bulge contributions exhibit an inner minimum with @xmath80 , while @xmath47 declines significantly as one moves further out . for low bulge masses , an additional outer
minimum emerges , within @xmath81 , before @xmath47 rises again outside the halo harmonic core .
only one of the @xmath47 curves , corresponding to the most massive bulge , has @xmath82 at _ all _ radii . therefore , for a given halo , there is a minimal bulge mass , @xmath83 , that is necessary to create sufficiently round loop orbits at all radii @xmath84 if we assume that the bulge core density varies at most by a factor of a few ( @xmath85 $ ] ) , @xmath86 , which is only weakly dependent on the bulge mass and density ( @xmath87 ) , varies little .
the minimal bulge mass is also nearly constant ( calculations show , for example , that it varies by @xmath88% when @xmath63 changes from 1 to 2.5 ) .
there is more sensitivity to the assumed @xmath18 . in fig .
[ complots ] ( asterisks ) we show the bulge mass required to create loop orbits with @xmath61 at all radii outside @xmath86 , for @xmath89 .
these are the masses necessary , at this density , to produce outer minima with the required values . for bulge densities
@xmath90 these minima lie at radii @xmath91 . thus , unless the sbh can contribute significantly to the potential at radii comparable to @xmath13 , which seems implausible , the dominant contributions to the potential at the radii examined here should be only those due to the bulge and halo components . therefore , once the halo parameters are fixed , the creation of loop orbits with given @xmath61 outside the bulge core will depend only on the bulge parameters @xmath86 and @xmath92 .
to2.5 in ' '' '' for @xmath89 , a well - defined inner minimum in the @xmath47 curve exists inside the bulge core only if @xmath93 . for larger @xmath18 , the inner minimum moves outward and merges with the outer one .
systems with these properties lie outside the scaling region in fig . 2 .
in this case , for the condition @xmath61 to be satisfied , the sbh has to contribute significantly at all radii no matter how massive the bulge is .
( for @xmath89 , the curve with @xmath94 in fig . 2 lies in this regime .
) furthermore , the minimal bulge density required to create loop orbits with @xmath61 , without appeal to an overmassive sbh , increases rapidly when @xmath95 ( fig .
[ exmags ] ) , suggesting that the bulk of the material may form stars before the condition @xmath61 is reached everywhere within the halo core .
if we demand that @xmath96 , then @xmath97 must be smaller than 0.8 in order for the model to be plausible , implying that @xmath40 must lie in the range @xmath98 .
to2.5 in ' '' '' on the other hand , star formation can be somewhat delayed to higher densities .
a reasonable range lies within @xmath99 .
still higher densities in the bulge can be excluded based on the observed rotation curves . in this range of @xmath63 ,
values of @xmath100 become feasible without invoking unrealistically massive sbhs , as seen from fig .
[ exmags ] .
one may also reasonably assume that , due to enhanced dissipation , higher densities require larger @xmath18 to maintain long - lived gaseous motion ( detailed modeling , however , will be necessary to determine exactly how @xmath18 depends on the system parameters ) .
in this situation , the range in @xmath40 that can be inferred from fig . 2 , is again about @xmath98 and compatible with the observed value of @xmath101 , which exhibits a significant scatter .
the values of @xmath40 obtained so far , which should be regarded as _ initial _ values , are fixed by the inner minimum of the axis ratio curves .
as argued in section 3.1 , further evolution of the bulge will occur unless sufficiently round loop orbits also exist at the outer minimum . in the next section
we examine how this evolution can modify the range of @xmath40 , if at all .
there are basic differences between the conditions for satisfying @xmath102 at the inner and outer minima in fig .
[ homofin ] . as discussed in section 3.1 , the initial growth of the sbh
is terminated after the inner loop orbits reach the critical eccentricity .
the growth of the bulge , on the other hand , continues until the required @xmath18 is reached at all radii inside the halo core .
since the increase of bulge mass , at roughly constant density , moves the inner minimum outward as described in section 3.1 , the sbh would have to continue growing in pace with the bulge mass if it is to maintain the critical value of @xmath40 ( which is determined only by the values of @xmath39 and @xmath18 ) .
does this actually happen , and if so , how ?
to2.5 in ' '' '' to see how the parallel growth of the sbh and bulge could come about , consider the solid curve in fig .
[ homofin2 ] , which represents the state of the sbh / bulge system at the end of the initial stage of infall .
this is equivalent to one of the curves ( say , the dash - dotted line ) in fig . 1 , except that we have chosen @xmath103 for the inner minimum , instead of 0.9 .
this line has two maxima : a left - hand one at the origin , determined by the sbh , and a right - hand one , determined by the bulge core .
note that the outer part of the solid curve drops below @xmath18 , implying that the bulge will continue to grow .
now suppose this growth occurs _ without _ corresponding growth of the sbh ( in contrast to the case in fig . 1 , where @xmath40 is kept constant ) .
if the bulge continues to grow while the sbh growth is stopped , the right - hand maximum moves further to the right , while the left - hand maximum stays the same .
this opens a widening ` gap ' at the position of the inner minimum which drops below @xmath18 , allowing gas to flow again toward the radius of influence of the sbh , @xmath104 .
this is illustrated by the dashed curve in fig .
[ homofin2 ] , which corresponds to a factor of 10 increase in @xmath45 accompanied by a similar _ drop _ in @xmath40 .
the gas is expected to accumulate in this vicinity .
however , once a substantial amount of gas has collected at @xmath104 , it becomes self - gravitating and is prone to global self - gravitating instabilities .
the fastest of these instabilities , @xmath105 modes or bar instability ( e.g. , bardeen 1975 ) , have been discussed in a similar context by shlosman , frank & begelman ( 1989 ) .
they induce rapid ( dynamical ) gas inflow .
hence we expect that the growth of the sbh at this stage will be intermittent , but because the time - averaged conditions for infall at the inner and outer maximum do not change , we expect the value of @xmath40 to stay roughly constant within the range given in fig . 2 , namely @xmath106 , depending on the value of @xmath18 .
is there always enough gas to accumulate at @xmath107 in order to trigger a bar instability , due to the opening of the gap between the sbh and the bulge ?
one can imagine the opposite extreme to that discussed above , in which insufficient gas enters the widening gap from outside , perhaps because star formation is efficient during the early stages of infall .
let us suppose , for the sake of argument , that the sbh does _ not _ grow beyond its initial mass , as determined by the initial value of @xmath40 .
how does this affect the range of final @xmath40 values ?
figure 4 shows minimal bulge masses , @xmath108 , for a range of @xmath18 values . while we have no estimate for the initial bulge masses which define the initial @xmath40 in fig .
2 , it is possible to rule out very small masses , e.g. , @xmath109 , because the total baryonic mass within the halo core is taken to be about 10% .
this means that for lower @xmath110 , the initial bulge mass is equal to or larger than @xmath83 estimated in section 3.3 , and , therefore , will not grow beyond its initial value .
it is easy to understand this result , because one needs small baryonic masses to create loop orbits with such large eccentricities .
this means that in this regime the initial value of @xmath111 ( in figures 2 and 4 ) is also its final value . alternatively ,
in the regime of larger @xmath112 , the bulge can grow at most by a factor of @xmath113 , reducing the initial @xmath40 by this amount . luckily , the initial values of @xmath40 for high @xmath18 lie around their high end , @xmath114 . reduction by up to two orders of magnitude brings them again to about @xmath115 .
hence , whether the sbh grows after the initial stage or not does not destroy the @xmath7 correlation and does not move the values of @xmath40 outside the observed range .
a corollary of this discussion is that the final value of @xmath40 appears insensitive to the value of @xmath18 , if the growth of the sbh is supressed , and depends strongly on @xmath18 , if the sbh grows in tandem with the bulge .
note also that the variation of @xmath116 ( represented by the asterisks in fig
. 4 ) , follows closely that of @xmath117 ( represented by filled circles ) . the ratio of these two quantities , @xmath118 is , therefore , largely independent of @xmath18 . in other words ,
the minimal bulge mass is predicted to be about 5 times the geometric mean between the sbh and the halo masses , provided that the sbh grows along with the bulge . to summarize ,
the initial value of the black hole - to - bulge mass ratio @xmath40 is fixed by the density at which the gaseous material in the inner region becomes self - gravitating and forms stars , and by the value of @xmath18 .
however , unless loop orbits of sufficient eccentricity are created in the _ whole _ harmonic halo core region , the mass of the bulge will continue to grow . at the same time
, the dynamical infall onto the sbh can choke , if the remaining gas mass is insufficient to cause bar instability and channel it to the sbh . in the latter case
@xmath40 will decline below its initial value as the bulge grows , but will tend to approach a final value roughly independent of @xmath18 . in the opposite limit , the sbh will grow in proportion to the bulge , and @xmath40 will stay constant ( as shown by figs . 2 and 4 ) . in either limit
, we assume that the process stops when the mass of the baryonic component reaches the minimal mass required to achieve loop orbits with a given @xmath61 , at all radii . for less triaxial haloes , and for a given bulge core density , lower sbh masses are required to produce loop orbits below a critical eccentricity .
the minimal mass of the bulge needed to create the required loop orbits in the outer region , however , also decreases especially since , for a mildly triaxial halo , the radius at which no loop orbits exist inside the harmonic core decreases with decreasing triaxiality .
thus , in the regime of mild halo triaxiality , the ratio of @xmath40 to ( normalized ) minimal bulge mass should not depend sensitively on the value of @xmath14 .
we do ignore the fact that triaxiality can be a function of radius , and choose a fixed value of @xmath14 for simplicity .
once orbits of sufficiently large @xmath47 are present , dissipation will reduce radial motion relative to these orbits , as well as vertical motion away from the symmetry plane defined by the angular momentum vector .
infalling gas will start to populate the newly - formed , round , non - intersecting periodic loops , leading to the formation of a disk component inside the halo core . at this stage
the bulge formation stops . outside the halo core
, gas can accumulate at any stage of the formation process on closed loop orbits , which always exist in strongly inhomogeneous density distributions .
provided that the halo triaxiality is small , these orbits will be nearly circular ( for potential axis ratio @xmath14 mildly deviating from unity @xmath119 : see , e.g. , rix 1995 ) . if the halo core radius is exceedingly large , however , it is possible that no significant disk component will form at all . for
if the core is large relative to the total halo ( virial ) radius , the contracting gaseous component will end up in the core , instead of spinning up and forming an extended disk .
losing most of its angular momentum to the halo , it will eventually end up as part of the bulge component .
this would lead to the formation of an elliptical rather than a disk - dominated galaxy . in this context , one expects that haloes with larger core radii are host to larger spheroidal components .
since it appears that , in general , more massive galaxies are usually of earlier type ( e.g. , persic , salucci & stel 1996 ) , one could deduce that more massive cores have larger core radii .
this would be expected if these cores followed a faber - jackson type relation ( for example , if @xmath120 , then @xmath121 ) as tentatively suggested by observations ( burkert 1995 ; salucci & burkert 2000 ; dalcanton & hogan 2001 ) and deduced if halo cores formed via the destruction of the inner ( @xmath122 ) regions of nfw haloes ( navarro , frenk & white 1997 ) . in this case , lower mass haloes form when the universe is denser and are , therefore , more concentrated . as a consequence , the region where @xmath122 is smaller relative to the virial radius for low mass haloes ( el - zant , shlosman & hoffman 2001 ) .
the existence of a faber - jackson type relation for halo cores is also required in order to reproduce the @xmath123 relation as discussed in the next section . within the above framework , the formation of the sbh is intimately tied to the bulge component , whereas the formation of the disk component takes place after the processes leading to bulge and sbh formation are essentially complete .
the bulk of the disk is also expected to form at scales larger than the halo harmonic core ( but see section 5 ) .
thus , the deduced correlation involving the bulge and sbh does not simply generalize to one involving the disk as well , in accordance with observations ( e.g. , kormendy & gebhardt 2001 ) .
the fact that all the orbital properties discussed in this paper depend only on the relative magnitudes of the masses and spatial scales of the galactic components involved has important consequences . for the @xmath7 relationship it has the obvious implication that the derived correlation will hold for all halo masses .
a more powerful prediction transpires in relation to the @xmath4 relation . for this to hold ,
it is necessary that the masses and velocity dispersions of halo cores are related in a similar manner .
suppose that for a given halo there exist unique values for the sbh and bulge masses , and for the bulge scalelength .
if a faber - jackson type relation between the halo core mass and velocity dispersion exists , a corresponding relation will exist between the bulge parameters .
it also follows that a similar relation will exist between the sbh mass and the bulge ( and halo ) velocity dispersion .
this can be deduced by simple scaling transformations , because the orbital properties we are interested in are all invariant with respect to spatial scale and mass transformations .
this means that if we multiply the masses of the sbh , bulge and halo by some constant @xmath124 , thus effectively changing the mass units , the curves in fig .
[ homofin ] will remain invariant . in the same manner , if we multiply all lengthscales by some factor @xmath125 , so that @xmath126 and @xmath127 , thus effectively changing the length unit , all curves in fig .
[ homofin ] remain the same .
this implies that if , for example , @xmath128 , then @xmath129 are equivalent systems in the sense decribed above ( they have the same axis ratio curves for their loop orbits with the same @xmath47 values as a function of rescaled radius ) .
these systems will all follow the @xmath3 relation for the halo . in this particular case
, @xmath130 implies a similar relationship between @xmath1 and @xmath2 , as well as @xmath0 .
it is of course possible that , for a given halo , the masses of the sbh and bulge , as well as the bulge lengthscale , are not unique . in other words , the subset of haloes with a given @xmath31 and @xmath13 may contain bulges and sbhs with a distribution of properties . in section 3.2 we had assumed that the bulge collapse is largely terminated when its density is of the order of the halo core density .
this in turn fixes @xmath40 , once @xmath18 is determined . in this section
we further assume that the value of @xmath40 is not changed as a result of any subsequent bulge growth .
that is , the sbh grows in tandem with the bulge ( cf .
section 3.4 ) .
this relaxes the assumption that the bulge masses are determined by the minimal mass required to create closed loops with @xmath61 at all radii , allowing for more massive bulges .
one would like to infer to what extent variations in the bulge and sbh properies , within a given halo , affect the homology relations discussed above .
we now show that when one accounts for variations of the bulge and the sbh parameters , the departure from the aforementioned relation is not dramatic .
this comes about basically because the mass and core radius of the bulge are correlated , under the assumptions of our model .
first , we note that the average velocity dispersion of the core of the baryonic component can be written as @xmath131 , and the average density @xmath132 . in this case @xmath133 . here
@xmath134 depends on the functional form of the density distribution .
its exact value is unimportant if all bulges are assumed to have the same functional form for their density distributions .
henceforth we set @xmath135 . for a constant bulge - to - halo density ratio @xmath78 a relationship between bulge velocity dispersion and mass @xmath136 results . in a toy model where the halo density does not vary at all with radius
a constant bulge density determines a unique @xmath40 , given @xmath18 , and an `` @xmath137 '' relation between bulge properties within a given halo arises , with index equal to 3 .
considering more realistic models for the density distribution raises the index somewhat .
when the core radius of the baryonic component is not very small compared to that of the halo , the density of the halo will not be strictly constant in the region of interest . as a result , the slope of the @xmath137 relationship will further increase . , where @xmath138 , the slope of @xmath123 is given by @xmath139 $ ] which stays between @xmath140 for @xmath141 , and then rises rapidly .
the resulting slope of @xmath123 is some weighted average of @xmath142 , and greater than 3 . ]
this occurs because larger bulges see " a smaller mean halo density , and therefore smaller bulge densities suffice to produce the same relative contribution to the potential . since @xmath143 , an inverse correlation between density and @xmath144 steepens the @xmath145 relation .
( a marginal effect is already seen in fig . 1 where the inner maxima produced by bulges with progressively larger @xmath146 also have progressively larger inner minima of @xmath47 . )
if the baryonic component giving rise to the gravitational potential is significantly flattened there will be an increase in the @xmath123 slope to 4 . in this case
, the surface , rather than the volume density , will determine the gravitational field , resulting in a situation where @xmath147 .
to2.5 in ' '' '' the fact that the index of the relation between @xmath2 and @xmath1 at constant bulge density is close to that of the faber - jackson relationship implies that , even if there is considerable variation in the bulge mass for a given halo , the departure from a faber - jackson relation defined by the halo parameters ( as described above ) is not too large .
this is illustrated in fig .
[ comsig ] , where we plot the relationship between the average velocity dispersion ( simply defined as @xmath148 ) of the baryonic component and its mass for systems having parameters @xmath40 and @xmath18 corresponding to those in fig .
[ compac ] .
the plots are obtained by keeping @xmath40 and @xmath18 constant and , for a given bulge mass , decreasing its characteristic radius until @xmath61 at all radii inside the halo core . to obtain physical parameters we set @xmath149 km s@xmath150 and @xmath151 kpc . the solid line in fig .
[ comsig ] is actually a superposition of several lines , corresponding to different values of @xmath18 ( and @xmath40 ) , as used in fig . 2 .
its average slope is around 3.2 and increases slowly toward higher dispersion velocities .
these lines , however , have different end - points , determined by the minimal bulge masses associated with the different @xmath18 ( cf . section 3.3 ) which are denoted by circles in fig [ comsig ] . as the mass decreases .
the result is that small hooks " ( not shown here ) appear in place of the circles in fig .
[ comsig ] . in principle , this regime may be of interest , but is not considered in this paper because it occurs for a very small range of bulge masses . ]
the maximum values of the bulge masses in these curves correspond to 2.5 times the halo core mass .
it seems unlikely that bulges would be more massive than this .
their characteristic densities or radii would have to be much larger than those of the halo , as would their contribution to the potential . in systems with disks
such extreme conditions would violate the shapes of observed rotation curves .
to2.5 in ' '' '' in reality , for a given halo , there is a range of possible bulge densities and associated values of the @xmath40 parameter .
this , in principle , affects the normalization of the @xmath145 relation .
nevertheless , the normalization turns out to be weakly dependent on @xmath78 ( @xmath152 ) .
thus , variations in bulge densities within haloes of given mass and size do not significantly affect the relationship between bulge mass and velocity dispersion . in our model ,
the central ( bulge ) density required for the production of loop orbits with @xmath61 in the scaling regime is also only weakly dependent on @xmath40 ( as @xmath153 , cf .
[ compac ] ) .
the above leads to an important corollary , _ that large variation in @xmath40 will cause only small changes in _ @xmath145 , as illustrated in fig .
[ varkmb ] .
this relation , therefore , appears to be robust and is not heavily affected by changes in bulge and sbh parameters .
note also that , for less dense bulges the slope of the lines in fig . [
varkmb ] tends to @xmath154 , especially in the limit of large values of the velocity dispersion ( which is also where most of the observations lie ) .
this is due to the the effect described above massive bulges with smaller densities are more extended , and , therefore probe larger regions of the halo core .
to2.5 in ' '' '' to obtain an @xmath4 relation , one has to multiply the values of the bulge masses in fig .
[ comsig ] by the appropriate @xmath40 factors , arriving at an @xmath4 relation _ within a given halo .
_ this introduces `` scatter '' by shifting the @xmath155 lines , as can be seen from fig .
[ varkmbh ] . for any fixed value of @xmath1
, this figure reveals that the ( vertical ) scatter in @xmath0 is similar to that in @xmath40 .
this leads us to an important point . in the present formulation ,
the @xmath145 relation is tighter than the @xmath4 one .
this appears to contradict observations suggesting that @xmath4 is much tighter than the faber - jackson relation .
however , this result has been obtained under the assumption that there is _ no scatter in the halo faber - jackson relation .
_ a significant scatter in @xmath3 would result in a corresponding scatter in @xmath156 , which follows from the homology scaling discussed in the beginning of this section .
moreover , if @xmath18 correlates with @xmath1 in such a way that gaseous sustems embedded in haloes with higher velocity dispersion require larger values of @xmath18 to remain stable , the scatter in the @xmath4 relation due to variations in @xmath40 can be reduced .
for example , in fig .
[ varkmbh ] , we have assumed a single halo for all the lines with different @xmath18 .
if , however , lines with higher values of @xmath18 are associated with haloes with larger @xmath157 ( and thus @xmath1 ) , the constant @xmath18 lines would be shifted in such a way that the vertical distances between them at a given value of @xmath1 decrease .
we illustrate this effect in fig [ shift ] , where a linear relationship is assumed between @xmath1 and @xmath18 .
to2.5 in ' '' '' we suggest that the relative tightness of @xmath4 could result from a loose relation between @xmath1 and @xmath18 , coupled with scatter in @xmath3 comparable to that of the faber - jackson relation for bulges .
such @xmath158 relation can result from a general trend that dissipation is increasing in more concentrated systems . testing
this will require detailed modeling of the way in which the dissipation rate depends on @xmath18 , velocity and density , and is outside the scope of this paper .
based on the illustrative example of fig .
[ shift ] , we only claim here a plausibility of a loose relationship between @xmath18 and the system velocity dispersion . if the postulated correlation persists for haloes of different masses , a steepening of the @xmath4 relation relative to the halo ( and bulge ) mass velocity dispersion relation is expected , implying that the former should have a larger index .
in the model presented here , gaseous baryonic material settles inside a mildly non - axisymmetric halo with a nearly constant density core . initially ,
no orbits with a definite sense of rotation exist .
the first infalling baryonic material , therefore , efficiently loses its angular momentum to the core .
this initial collapse terminates only when the gas becomes self - gravitating and forms stars . the increased central density concentration which is produced in this first phase , however , destroys the harmonic core , paving the way for the existence of non - intersecting closed loops with a definite sense of rotation .
if the loop orbits are too eccentric , the gas will shock and depopulate them .
these orbits , therefore , can not represent long - lived attractors of the dissipative motion .
thus , if the baryonic component does not possess a central cusp initially , the absence of sufficiently round supporting closed loops will lead to the formation of a central mass concentration including an sbh . at larger radii
the potential responsible for creating sufficiently round loop orbits is that of the extended baryonic component , in the form of a bulge .
this leads to a linear relationship between the bulge and sbh masses .
it is crucial that both the onset of self - gravity and the appearance of increasingly circular loop orbits are subject to the same condition , both depending on the density ratio of the collapsing gas to that of the background halo core .
this limits the allowable range in densities . for bulge
core densities of order those of the halo core , and for plausible values of the critical eccentricity , the value of @xmath40 lies in the range @xmath98 , which is compatible with present observations . for densities up to an order of magnitude larger , the range of values is the same , provided that the critical values of the closed loop axis ratios are larger .
this can be expected if the dissipation rate along loop orbits is dependent on density a plausible assumption .
the most important prediction of the model outlined above is that the bulge and sbh parameters are determined by the halo properties .
in particular , relationships between sbh and bulge masses and the velocity dispersion of the bulge necessarily arise , and with the right exponents , if the haloes should also exhibit a faber - jackson type relation between their masses and velocity dispersions .
moreover , within a given halo , variations in the bulge and sbh properties are not expected to destroy these relations because imposing the condition of critical eccentricity requires that the bulge and sbh masses are related to the bulge velocity dispersion _ via _ a power law , with index also close to that of the faber - jackson relationship . finally , if the faber - jackson relationship for the halo exhibits significant scatter and if , as again seems plausible , the critical eccentricities of the loops anti - correlate with the density of the system , less scatter should be present in the relationships between sbh masses and bulge velocity dispersions than the corresponding relationships between bulges masses and their own velocity dispersions the standard faber - jackson relationship .
there is already tentative evidence that halo cores may indeed follow faber - jackson type relationships ( burkert 1995 ; dalcanton & hogan 2001 ) .
in addition , a halo core produced by flattening out the inner region ( where @xmath8 ) of the nfw profile would produce such a relation ( el - zant , shlosman & hoffman 2001 ) .
the model thus makes testable predictions concerning the relationships among the very inner regions of galaxies , their extended baryonic components , the dark matter haloes they are thought to be embedded in and the cosmology that predicts their existence . in this framework ,
larger and more massive halo cores produce , on average , larger and more massive bulges .
disks form outside the core , or later , when the central concentration produced by the baryons destroys the core .
if the core is very large , most of the baryonic material is consumed in the first phase and no significant disk forms .
this effect is expected to be prominent in larger mass cores , since if these follow the faber - jackson relation , more massive haloes should have proportionally larger cores .
other predictions include the requirement that the average density of the bulge in the central region should be close to that of the halo core .
there is also a minimal bulge mass associated with a given core , although this varies significantly with the critical loop orbit eccentricity assumed .
a number of additional consequences for galaxy formation and evolution will be discussed elsewhere . to obtain relationships between sbh masses and bulge properties ,
we have assumed that during the gaseous infall phase the baryonic component did not have a central density cusp . most observed bulges and ellipticals , however , do have such cusps .
one then has to assume that once star formation starts , cold dissipational collapse is initiated .
this would lead to a central density cusp , as it does in the case of cosmological haloes .
memory of the initial state is retained via the total energy , which determines the final velocity dispersion .
merritt & ferrarese ( 2001 ) have suggested that a `` self - regulating '' mechanism , related to the threshold mass necessary for the loss of triaxiality in the system , may be behind the close correlation between sbh and bulge properties . as these authors point out , however , the sbh masses required for strong chaotic behavior leading to rapid loss of triaxiality are in fact probably too large being of the order of a few percent of the mass of the system s baryonic component .
this is actually of the order of the sbh mass needed to create round loop orbits at _ all _ radii inside the halo core . in our model , however , this is not assumed . instead , an additional baryonic component plays the role of creating these orbits in the outer region . the collusion between this `` bulge '' component and the sbh in destroying the harmonic core and creating a situation whereby stable gaseous motion can exist gives rise to the correlations described in this paper .
the correlations obtained here are compatible with the observed ones , and with acceptable scatter , despite our lack of knowledge of the values of such parameters as @xmath18 and its variation with system properties .
detailed modeling of the gas dynamics will be required to further constrain this model .
it is also possible that our distinction between the dynamical role played by the sbh and that played by the bulge core is too restrictive .
we have introduced this to be able to obtain quantitative results , within the model , solely on the basis of the orbital characterisitics . in general , the roles of the two components , i.e. , the sbh and the bulge , may not be too distinct formation of the sbh can take place simultaneously with a cuspy bulge . for this , the even more ambitious task of a self - consistent treatment , including gas and stellar dynamics and star formation , is required .
we believe , however , that our results are generic and arise from fundamental dynamical phenomena which will manifest themselves in any formulation of galaxy formation in mildly triaxial haloes with harmonic cores .
we thank our colleagues , too numerous to mention , for stimulating discussions .
this work was supported in part by nasa grants nag 5 - 10823 , wku-522762 - 98 - 6 and hst go-08123.01 - 97a ( provided by nasa through a grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 ) to i.s
. , nsf grant ast-9876887 to m.c.b . , and nsf grant ast-9720771 to j.f .
arnold , v.i .
1989 , mathematical methods of classical mechanics ( new york : springer ) bardeen , j.m .
1975 , proc .
a.hayli ( dordrecht : reidel ) , 297 binney , j. , tremaine , s. 1987 , galactic structure , princeton univ .
press bogoliubov , n.n . ,
mitropolsky , y.a .
1961 , asymptotic methods in the theory of non - linear oscillations ( new york : gordon & breach ) burkert , a. 1995 , , 447 , l25 cole , s. , lacey , c. 1996 , mnras , 281 , 716 dalcanton , j.j . , hogan , c.j .
2001 , , 561 , 35 de blok , w.j.g . , bosma , a.c .
2002 , a&a , 385 , 816 de zeeuw , t. , merritt , d. 1983 , , 267 , 571 ( dzm ) dubinsky , j. 1994 , , 431 , 617 el - zant , a. 1999 , phys rep , 311 , 279 el - zant , a. , shlosman , i. , hoffman , y. 2001 , , 560 , 636 faber , s.m . , jackson , r.e .
1976 , , 204 , 668 ferrarese , l. , merritt , d. 2000 , , 539 , l9 frank , j. , king , a. , raine , d. 2002 , accretion power in astrophysics , cambridge univ . press , 3rd edition gebhardt , k. , et al .
2000 , , 539 , l13 heller , c.h . , shlosman , i. 1994 , , 424 , 84 kormendy , j. , gebhardt , k. 2001 , 20th texas symp . on relativistic astrophysics , j.c .
wheeler & h. martel , eds .
( aip conf .
586 ) , 363 kuijken , k. , tremaine , s. 1994 , , 471 , 178 locas , e.l . ,
hoffman , y. 2000 , , 542 , l139 merritt , d. , ferrarese , l. 2001 , the central kpc of starbursts & agn : the la palma connection , j.h .
knapen , j.e .
beckman , j.h .
, i. shlosman & t.j .
mahoney , eds .
( asp : san - francisco , vol 249 ) , 335 miyamoto , m. , nagai , r. 1975 , , 27 , 533 navarro , j.f .
, frenk , c.s . , white , s.d.m .
1997 , , 490 , 493 persic , m. , salucci , p. , stel , f. , 1996 , , 281 , 27 pfennifer , d. , norman , c.a .
1990 , , 363 , 391 rix , h .- w . , 1995 , iau symp .
169 , unsolved problems of the milky way , astro - ph/9501068 rix , h .- w . ,
zaritsky , d. 1995 , , 447 , 82 salucci , p. , burkert , a. , 2000 , , 337 , l9 shlosman , i. 2002 , agns : from central engine to the host galaxy , s. collin et al .
( asp conf .
series , vol .
290 ) , in press shlosman , i. , frank , j. , begelman , m.c .
1989 , nature , 338 , 45 tremaine , s. , et al .
2002 , astro - ph/0203468
let the potential @xmath159 be a function of the cartesian coordinates in the halo symmetry plane , @xmath160 with @xmath161 .
we follow de zeeuw & merritt ( 1983 , hereafter dzm ) and expand the potential in even powers of the coordinates : @xmath162 if no bulge exists , inside the halo s harmonic core only the terms quadratic in the coordinates are important . if , due to an additional component , or at the boundaries of the core , weak nonlinearity is present the series can be truncated , as above , to second order , as higher order terms are unimportant .
the coefficients are given by : @xmath163 , @xmath164 , @xmath165 , @xmath166 , @xmath167 , where the derivatives are taken at @xmath168 . in terms of these one defines the auxilliary variables @xmath169 , @xmath170 , @xmath171 .
the condition for stability of the loop orbits is ( see dzm ; table 2a third row ) @xmath172 we will be interested in systems that are both mildly nonlinear and mildly nonaxisymmetric : thus @xmath173 and @xmath174 .
the above condition reduces to @xmath175 which , under the above conditions , is always satisfied .
it is , therefore , the condition for the existence of loop orbits and not their stability that will be of interest to us . in the absence of a central mass ( sbh ) or density cusp these can not be found arbitrarily close to the center , instead there is a _ bifurcation radius _ beyond which these exist .
this determines the effective core radius of the system . to second order
, the condition for the existence of loop orbits is @xmath176 with @xmath177 and ( to first order ) @xmath178 , where @xmath179 is the hamiltonian . for @xmath180 ( first order
potential terms dominate in eq . [ pot ] ) and @xmath181 , the above requires @xmath182 in the unperturbed case , the action variables ( e.g. , binney & tremaine 1987 ) are time - independent and ( exact ) solutions can be written in terms of these in cartesian coordinates as @xmath183 , etc . in the mildly nonlinear case the solutions of the equations of motion , averaged over a dynamical time ( denoted below by a bar " ) , approximate the true solution to first order in the ( relative amplitude of the ) perturbation and for a number of dynamical times inversely proportional to this ( dzm ; see also bogoliubov & mitropolsky 1961 ; arnold 1989 ) . in this case
analogous approximate solutions can be given in terms of the corresponding action variables .
in particular , for loop orbits in mildly nonlinear potentials one finds @xmath184 and @xmath185 eliminating @xmath186 one gets @xmath187 where @xmath188 , @xmath189 and @xmath190 . at bifurcation , to first order , @xmath191 where @xmath179 is given by eq .
( [ hamiltonian ] ) .
these orbits are infinitely thin and represent oscillations along the @xmath192-axis , with amplitude @xmath193 which is the effective core of the bulge - halo system , as mentioned above . as one increases @xmath179 these become thicker with axis ratio @xmath194
the view we have taken in this paper is that when this ratio becomes large enough , @xmath195 , such loop orbits can support gaseous motion , and that stars formed on these orbits can constitute populations of stellar disks , thus ending the bulge formation stage .
we note here that , beyond the bifurcation point , the above relation predicts a rather rapid ( as a function of radius ) transition to round loop orbits .
this is confirmed by orbital integration , even though if the bulge mass is smaller than a certain minimal mass , in the fully nonlinear treatement , the orbital axis ratio can decrease again .
the second minimum in axis ratio curves ( e.g. , fig .
1 ) is thus not reproduced by this perturbation analysis .
we now apply the perturbation analysis to the superposition of potentials given by eqs .
( [ logarithm ] ) and ( [ bulgepot ] ) with the goal of calculating the bifurcation radii of the loops .
this is the effective harmonic core radius of the bulge - halo system .
since we are only interested in orbit shapes , we set @xmath196 and use scaled variables , as defined in section 2 . for the halo ,
we use @xmath197 where @xmath198 , corresponding to @xmath199 in eq .
( [ logarithm ] ) , parametrizes the nonaxisymmetry . for the bulge , we have @xmath200 with the above definitions for the potential we have @xmath201 @xmath202 @xmath203 and @xmath204 from equations ( [ hamiltonian ] ) , ( [ action ] ) , and ( [ ymax ] ) bifurcation happens when the value of the long axis of the loops satisfies @xmath205 substituting from the expressions above and taking the limits @xmath206 , @xmath207 , and assuming @xmath208 , we obtain @xmath209 this is the effective harmonic core radius of the bulge - halo system . | the masses of supermassive black holes ( sbhs ) show correlations with bulge properties in disk and elliptical galaxies .
we study the formation of galactic structure within flat - core _ triaxial _ haloes and show that these correlations can be understood within the framework of a baryonic component modifying the orbital structure in the underlying potential . in particular , we find that terminal properties of bulges and their central sbhs are constrained by the destruction of box orbits in the harmonic cores of dark haloes and the emergence of progressively less eccentric loop orbits there .
sbh masses , @xmath0 , should exhibit a tighter correlation with bulge velocity dispersions , @xmath1 , than with bulge masses , @xmath2 , in accord with observations , if there is a significant scatter in the @xmath3 relation for the halo . in the context of this model
the observed @xmath4 relation implies that haloes should exhibit a faber - jackson type relationship between their masses and velocity dispersions .
the most important prediction of our model is that halo properties determine the bulge and sbh parameters .
the model also has important implications for galactic morphology and the process of disk formation .
3mark iii # 1#2#3#4=.24 = .24 = .24 = .24 |
the spin-1/2 of a single electron trapped in a quantum dot ( qd ) is a promising candidate for a carrier of quantum information in a quantum computer @xcite . to perform a quantum computation
we need to have all the unitary operations from some universal set of quantum gates at our disposal @xcite .
one such universal set consists of all the single qubit quantum gates and a two - qubit controlled - not ( cnot ) quantum gate .
quantum computation over the single - spin qubits with the logical states corresponding to the spin orientations @xmath0 and @xmath1 can in principle be achieved using an external magnetic field or with g - factor engineering for the single qubit operations , and with the time - dependent isotropic exchange interaction @xmath2 for manipulating a pair of qubits encoded into spins @xmath3 and @xmath4 @xcite .
control of electron spins in quantum dots is in the focus of many intense experimental investigations .
manipulation of pairs of electron spins using the tunable isotropic exchange interaction has already been demonstrated in several experiments @xcite .
such control was used in a study of the qd spin decoherence due to the hyperfine coupling to the surrounding nuclear spins , where the splitting between the singlet states with the total spin @xmath5 and the triplet states with @xmath6 was used to turn on and off the singlet - triplet mixing caused by the hyperfine interaction .
an important result of these studies is that the coherence time of an electron spin in a quantum dot is very long if the decoherence due to the interaction with the nuclear spins can be suppressed .
the spin coherence times can be improved by the manipulation of nuclear spins @xcite , in principle allowing for elaborate sequences of operations to be performed .
single spin control is based on the local manipulation of the magnetic field or g - factor @xcite , or on esr methods @xcite and has only recently been demonstrated experimentally @xcite .
the difficulty of single - spin control has inspired a number of proposals for quantum computation based on the encoding of qubits into more than one spin .
these encoding schemes reduce the requirement on the control over electron spins , but have the drawback of introducing so - called leakage errors in which the state of encoded qubit `` leaks '' out of the set of computational states .
standard error - correction procedures can be modified to prevent this kind of errors @xcite
. a universal set of quantum gates operating on qubits encoded into states of three quantum dot spins with equal total spin quantum numbers can be implemented through control of the isotropic exchange coupling alone @xmath7 @xcite .
control over interactions that are symmetric only with respect to rotations about a fixed axis in spin space allows for the construction of a universal set of quantum gates that operate over qubits encoded into pair of spins .
one such encoding is into the orthogonal states @xmath8 and @xmath9 of two spins-1/2 .
a universal set of quantum gates over such qubits can in principle be performed by the control over @xmath7 , with the anisotropy provided by an external static homogeneous magnetic field and a site - dependent @xmath10-factor @xcite .
we consider a variant of the two - spin encoding where the logical zero @xmath11 and the logical one @xmath12 quantum states are the singlet and the triplet with zero projection of the total spin to the symmetry axis @xmath13 ( @xmath14 ) , e.g. , for lateral qds , the @xmath13 axis is the normal to the plane of the heterostructure , @xmath15 these qubits can be manipulated by an axially symmetric interaction to produce a universal set of quantum gates . the interaction with an external magnetic field and the isotropic exchange @xcite , or the interaction with an external magnetic field and an anisotropic spin - orbit coupling @xcite , or the spin - orbit coupling alone @xcite , were all proposed as a way of producing a universal set of quantum gates operating on singlet - triplet two - spin qubit , eq .
( [ eq : stqubit ] ) .
recently , it was suggested that an architecture based on singlet - triplet qubits individually addressed using isotropic exchange interaction and inhomogeneous magnetic field and coupled through coulomb interaction of the electrons is scalable and in principle realizable @xcite . in this paper , we study a particular realization of entangling two - qubit gates between singlet - triplet qubits , eq .
( [ eq : stqubit ] ) , where each qubit is represented by a pair of tunnel - coupled single - electron quantum dots , as proposed in @xcite . in this realization ,
the double quantum dots are separated by a barrier which is impenetrable for the electrons , so that the qubits are coupled exclusively through the coulomb repulsion of electrons , while the exchange terms between electrons on different double quantum dots vanish .
the setup of this double - double quantum dot ( ddqd ) is illustrated in fig .
[ fig:4dots ] . encode a qubit .
two such double quantum dot ( dqd ) qubits , left ( l ) and right ( r ) , at the distance @xmath16 are separated by an impenetrable barrier .
the tunneling matrix element @xmath17 within the double quantum dots ( dqd ) carrying the qubits , and the bias @xmath18 of the inner dots with respect to the outer are equal on both dqds , and can be electrically tuned . the coulomb interaction between the dqd
is represented by the capacitor @xmath19.[fig:4dots],width=321 ] the coulomb interaction is spin - independent , leading to an isotropic interaction @xmath20 between tunnel - coupled spins @xmath3 and @xmath4 . the anisotropic correction to this interaction is dominated by the spin - orbit coupling induced term @xmath21 .
the relative strength of the anisotropic interaction in the quantum dot systems in gaas is estimated to be @xmath22 @xcite .
the influence of the anisotropic corrections can be reduced in specific implementations of the quantum gates @xcite . in our study of a two - qubit gate operation
, we will only consider the case of isotropic interaction and neglect the weak anisotropy . in this case , transitions between spin - singlet and spin - triplet states on a dqd are forbidden . due to this spin symmetry , the four - electron hamiltonian is block - diagonal , @xmath23 the non - zero blocks @xmath24 , where @xmath25 , act on the states in which electron pairs on each dqd are either in the singlet ( @xmath26 ) or in a triplet ( @xmath27 ) state of the total spin @xmath5 or @xmath6 .
our main results are the effective low - energy spin interaction and a scheme to perform a two - qubit cnot gate in an electrically controlled ddqd system .
the effective low - energy spin interaction in this setup has the form @xmath28 two pairs of spins , @xmath29 and @xmath30 on the left @xmath31 qubit and @xmath32 and @xmath33 on the right @xmath34 qubit ( see fig . [ fig:4dots ] ) interact via the isotropic exchange interaction of strength @xmath35 , and the entangling interaction of strength @xmath36 that shifts the energy of the singlet - singlet state .
we show how the entangling two - qubit quantum gates for universal quantum computation can be performed through the electrical control of @xmath36 .
the triplet states with @xmath37 are degenerate in the absence of a magnetic field . a uniform magnetic field
@xmath38 , pointing along the @xmath13 axis normal to the plane of qds causes a zeeman splitting @xmath39 between the @xmath14 states and the states with @xmath6 , @xmath40 .
our results apply both to the isotropic ( @xmath41 ) and anisotropic , but axially symmetric ( @xmath42 ) case , if we take the @xmath14 state to represent the qubit @xmath43 state .
a two - qubit quantum gate can in principle be performed by adiabatically varying the tunneling matrix element @xmath17 and the bias @xmath18 within the dqd . in practice , it is much simpler to change the bias @xmath18 while @xmath17 remains fixed @xcite .
the control parameters @xmath18 and @xmath17 have to vary slowly on the time scale set by the energy splitting between the states of a given spin configuration . during the gate application ,
the orbital components of the @xmath26 and @xmath27 states are different due to the pauli principle that forbids the electrons in a spin triplet to share their orbital state , see fig .
[ fig : dotpulse ] . )
the ground state is the doubly occupied inner dot .
due to the pauli principle , only the spin singlets @xmath44 can tunnel into the doubly occupied states on their dqds .
as the bias @xmath18 is reduced , the states again become degenerate .
b ) a quantum gate is performed by sending a bias pulse @xmath45 . each qubit state @xmath46 acquires a phase @xmath47 , where @xmath48 is the ground state energy of the hamiltonian at time @xmath49 reduced to the appropriate spin subspace , resulting in a two - qubit quantum gate .
[ fig : dotpulse],width=321 ] as opposed to @xmath17 and @xmath18 that are determined by gate voltages and can be changed more or less at will , the coulomb interaction is set by the geometry of the system and therefore fixed .
we show how the control of the parameters @xmath17 and @xmath18 , or even @xmath18 alone , can nevertheless be used to implement entangling two - qubit gates on encoded singlet - triplet qubits through its influence on the coulomb terms . when an adiabatic gate is applied , the lowest energy state in each block @xmath24 , of energy @xmath50 , where @xmath25 , see eq .
( [ eq : hblocks ] ) acquires a phase @xmath51 .
the energy @xmath50 becomes time - dependent through the time - dependence of the parameters @xmath17 and @xmath18 in the interval @xmath52 .
the resulting interaction is described by an effective 4-dimensional two - qubit hamiltonian acting in the space spanned by the lowest - energy states @xmath53 , @xmath54 , @xmath55 , and @xmath56 in the corresponding blocks @xmath24 , and has the form of eq .
( [ eq : hstrongbias ] ) . in the regime of strong bias , @xmath57 , where @xmath58 is the on - site coulomb repulsion
, we investigate the ddqd system using perturbation theory . for the case of arbitrary bias @xmath18 , we numerically diagonalize the hamiltonian eq .
( [ eq : hblocks ] ) .
we show that the two - qubit quantum gate can be operated by tuning the bias @xmath18 so that the amplitude of the doubly occupied state in the lowest energy spin singlet becomes appreciable . in this `` on '' state with a large double occupancy amplitude ,
entanglement is generated between the two - spin qubits .
the entanglement generation is suppressed in the `` off '' regime with weak bias and tunneling .
therefore , the generation of entanglement between the two two - spin qubits encoded into ddqd can be efficiently controlled using the bias @xmath18 alone .
together with the single - qubit operation this control is sufficient for universal quantum computing .
this paper is organized as follows . in sec .
ii , we introduce our model of the ddqd system , followed by the discussion of the control through voltage pulses . in sec .
iii we focus on the case of the strongly biased ( @xmath59 ) ddqd system and calculate the interaction between the qubits .
the constraint of the strong bias is lifted in sec .
iv , where we numerically find the interaction between the qubits , valid at an arbitrary bias @xmath18 . in sec .
v , we outline the construction of a cnot gate based on the resources for the control over a pair of qubits deduced from the results of secs .
iii and iv .
our results are summarized in sec .
the technical details of the calculation are collected in the appendix a.
for the purpose of finding the effective low - energy spin hamiltonian , the excited orbital states of single quantum dots can be neglected , leading to the hund - mulliken ( hm ) approximation with one orbital per dot @xcite . in the hm approximation ,
the state space of the two - electron system in a double quantum dot ( dqd ) encoding the left ( @xmath60 ) or the right ( @xmath61 ) qubit is spanned by three singlet basis states , @xmath62 , @xmath63 and @xmath64 and one triplet basis state @xmath65 @xmath66 where @xmath67 is the annihilation operator for an electron in the state @xmath68 on the qubit @xmath69 , with position @xmath70 , where @xmath71 stands for inner and @xmath72 for outer quantum dot within a qubit , and spin @xmath73 .
the vacuum @xmath74 is the state of empty qds . in the standard notation
the singlet states of a dqd are denoted by @xmath75 , where @xmath76 is the number of electrons on the left qd and @xmath77 is the number of electrons on the right qd . our singly occupied singlet is then expressed as latexmath:[$|{\bar s}\rangle\equiv @xmath60 , dqd are @xmath79 , and @xmath80 . on the right , @xmath61 , dqd the definitions are reversed , @xmath81 , and @xmath82 .
the orbital states annihilated by @xmath67 approximate the ground states of the single - particle hamiltonian @xmath83 describing an electron in the magnetic field @xmath84 and confined to the system of quantum dots by the electrostatic potential @xmath85 .
the quantum dots form in the minima of this potential , which is locally harmonic with the frequency @xmath86 .
the ground states of @xmath87 localized in these wells are the translated fock - darwin states @xcite .
the hm hamiltonian is of the generic form @xmath88 the intra - dqd tunneling term @xmath89 preserves the electron spin .
the bias @xmath18 of the inner ( @xmath90 ) qds with respect to outer ( @xmath91 ) is taken to be symmetric , i.e. , the energy of both inner dots is lowered by the same amount .
the two - body coulomb interaction is denoted by @xmath92 .
near the center of the quantum dot , the electrostatic potential is approximately harmonic and we assume that the wave functions of the electrons annihilated by the operators @xmath67 are well approximated by the orthogonalized fock - darwin ground states . the impenetrable barrier that separates the dqds imposes the conservation of the number of @xmath93 electrons , @xmath94 , where @xmath95 .
the @xmath96 conserving terms , proportional to the interaction matrix elements @xmath97 in eq .
( [ eq : full_h ] ) , where the indices @xmath98 denote the single qd ground states , can be divided into intra - dqd terms where @xmath99 and inter - dqd terms that satisfy @xmath100 and @xmath101 .
all the other terms , e.g. , the ones that annihilate two electrons on the left ( @xmath102 ) dqd and create two on the right ( @xmath103 ) dqd violate the conservation of the electron numbers and therefore vanish .
the terms for the interaction within a dqd in eq .
( [ eq : full_h ] ) were discussed in @xcite .
they renormalize the one - body tunneling matrix element @xmath104 , introduce the on - site repulsion @xmath105 of two electrons on the same qd , and cause transitions between the two doubly occupied dqd states with the matrix element @xmath106 .
also , the coulomb interaction on a dqd contributes @xmath107 to the electrostatic energy of the symmetric and @xmath108 to the antisymmetric singly occupied orbitals of two electrons in a dqd @xcite , giving rise to a direct exchange interaction between spins . as a result ,
the electrons on a dqd are described by an extended hubbard model with the isotropic exchange interaction @xcite @xmath109 where @xmath110 is the effective on - site repulsion .
the coulomb interaction between the dqds produces three new classes of direct terms in the hamiltonian , while the exchange terms between the dqd vanish due to the impenetrable barrier . in the first class
are the terms proportional to the number operators @xmath111 , describing the electrostatic repulsion of the electrons in states @xmath112 and @xmath113 , where @xmath114 and @xmath115 . for a pair of identical dqds , there are three such terms : the interaction of a pair of electrons on the inner qds , @xmath116 , the interaction of an electron on the inner qd of one dqd and an electron in the outer qd of the other dqd , @xmath117 , and the interaction of electrons on the outer qds , @xmath118 , fig .
[ fig : correlation]a . in the second class
are the terms proportional to @xmath119 , where @xmath120 and @xmath121 .
these terms describe the spin - independent correction to the tunneling matrix element in the @xmath122 qubit due to the interaction with an electron in the state @xmath112 .
the two parameters that determine the tunneling corrections are @xmath123 , and are due to the interaction with an electron in the @xmath124 orbital in the other dqd , fig .
[ fig : correlation]b .
the terms in the third class are proportional to @xmath125 , and describe the processes in which electrons in both dqd tunnel simultaneously , fig .
[ fig : correlation]c .
the two independent matrix elements for these processes are @xmath126 describing the tunneling from the inner to the outer orbital in one dqd and from the outer to the inner in the other , and @xmath127 describing the simultaneous tunneling into inner or outer orbitals in both dqds . for the system in zero magnetic field these two matrix elements
are equal , @xmath128 . ) , at a medium distance ( @xmath129 ) or far apart ( @xmath130 ) .
b ) the tunneling matrix elements within a dqd are renormalized by @xmath131 or @xmath132 , due to the interaction with an electron on the inner or the outer dot of the other dqd .
c ) the interaction enables the correlated hopping processes in which electrons simultaneously tunnel in both dqds .
in one such process the electrons tunnel to the same side ( either left or right ) with the matrix element @xmath133 .
in the other correlated hopping process electrons simultaneously tunnel into the inner or outer quantum dots of their double quantum dots with the matrix element @xmath134 .
[ fig : correlation],width=321 ] in order to describe the influence of the intra - dqd tunneling @xmath17 and the bias @xmath18 on the spectrum of the ddqd , we have to model the dependence of the hamiltonian on these external parameters . in an experiment , both @xmath17 and @xmath18
are controlled by applying voltages to the electrodes that define the quantum dots .
the exact form of the voltage - dependent ddqd binding potential was studied using the schrdinger - poisson equation @xcite , but here we do not attempt to calculate the dependence of the hamiltonian eq .
( [ eq : hblocks ] ) on @xmath18 and @xmath17 from first principles . instead
, we adopt a quartic double - well model for the potential of a dqd centered at @xmath135 of the form @xcite @xmath136 where @xmath77 is the electron effective mass , @xmath137 is the distance between the approximately harmonic wells in a dqd , and @xmath16 is the distance between the dqd double - well minima . in the limit of well separated dots , @xmath138 , where @xmath139 is the qd bohr radius given by @xmath140 , and near the local minima of the quartic potential well at @xmath141 , the potential is approximately harmonic with the frequency @xmath86 .
the fock - darwin ground state wave functions in this harmonic potential centered at @xmath142 and in the magnetic field @xmath143 normal to the plane of the dots , described in the symmetric gauge by the vector potential @xmath144 , are @xmath145 where @xmath146 is the electron larmor frequency and @xmath147 is the resulting confinement frequency with both electrostatic and magnetic contributions .
we will use the magnetic compression factor @xmath148 to measure the strength of the magnetic field , consistently with the notation in @xcite . the translated single - electron fock - darwin states @xmath149 define the state space of the variational hm approximation for a ddqd .
the tunneling matrix element between the fock - darwin ground states in the local minima of the potential eq .
( [ eq : doublewell ] ) is our control parameter @xmath17 @xcite , @xmath150 where @xmath151 , is the overlap between the fock - darwin ground states in a dqd . as @xmath17
is changed by external voltages , we assume that the overlap @xmath26 between the oscillator states remains consistent with the relation eq .
( [ eq : todb ] ) which is valid for the double - well potential @xmath85 .
all the coulomb matrix elements can be expressed in terms of @xmath26 so that after solving equation ( [ eq : todb ] ) for the overlap they become functions of @xmath17 , see appendix a. the bias @xmath18 is modeled as an energy shift of the orbitals , so that the inner @xmath90 orbitals have their energy reduced by @xmath18 .
the two - qubit gates are applied by time - dependent tuning of the tunneling matrix element @xmath17 and/or the bias @xmath18 in the dqds using voltage pulses . in a typical experiment ,
the control of the qd energies through @xmath18 is much easier to achieve than the control over tunneling matrix element @xmath17 @xcite .
the reason behind this is that the energy bias is linear in applied voltage , while the tunneling is typically exponential .
the structure of the energy levels is particularly simple in the limit of zero tunneling @xmath152 . in this limit ,
the eigenstates are the hund - mulliken basis states , eqs .
( [ eq : states])([eq : statesend ] ) .
their energies are determined by the bias @xmath18 , the external magnetic field @xmath143 , and the direct coulomb interaction that is set by the device geometry .
a drastic change in the structure of the ddqd spectrum as a function of bias @xmath18 appears at the crossings of the lowest energy singlet states within a dqd .
each of the singlet states @xmath153 , @xmath63 , and @xmath64 is lowest in energy for some values of the bias @xmath18 , fig .
[ fig : flow ] . .
the energy of the singlet state with doubly occupied outer quantum dot , @xmath64 , is independent of the bias .
the energies of the singly occupied singlet , @xmath154 , and the singly occupied triplet , @xmath65 , state are lowered with the increasing bias as they have a contribution @xmath155 from the biased inner quantum dots .
the energy of the singlet with doubly occupied inner quantum dots , @xmath63 , is lowered with the increasing bias faster than the energy of @xmath154 and @xmath65 state , due to the bias contribution of @xmath156 . when the tunneling @xmath17 is zero , the lowest energy levels cross at the bias @xmath157 , leading to a drastic change of the effective spin interaction .
for nonzero tunneling , the levels anticross , but the effective spin interaction still changes significantly when we tune the system from one side of the anticrossing to the other .
[ fig : flow],width=321 ] a crossing occurs when either the positive bias overcomes the effective on - site repulsion @xmath158 , making the state with both electrons in an inner dot @xmath63 the lowest in energy , or the negative bias makes @xmath64 the lowest in energy , see fig .
[ fig : gndbias ] .
we use the effective on - site repulsion @xmath158 to emphasize the fact that it includes not only the repulsion of two electron in the same dot , denoted by @xmath58 , but also the energy of the interaction with the electrons on the other dqd .
we will also use two special values of the effective on - site repulsion , @xmath159 . due to the dependence of the effective on - site repulsion on the state of the other dqd , the lowest energy singlet - singlet ddqd state can consist of different singlets on the two dots , as in @xmath160 and @xmath161 . in the strong bias regions , the lowest energy singlets are doubly occupied states . for @xmath162
the lowest energy singlet is @xmath163 , and for @xmath164 , the lowest energy singlet is @xmath165 .
the second doubly occupied singlet state is separated by an energy gap @xmath166 from the lowest energy state .
of the inner quantum dots with respect to the outer ones is weaker than the effective on - site coulomb repulsion @xmath158 , the charge configurations of the lowest energy singlet and triplet states consists of singly occupied orbitals .
( b ) when @xmath167 , the lowest energy singlet has a doubly occupied inner quantum dot , while the orbital state of the lowest energy triplet remains unchanged .
[ fig : gndbias],width=321 ]
to develop an intuitive picture of the operation of an entangling two - qubit gate and the mechanisms for its control , we consider the simple case of strong bias .
we show how the switching between the strong bias regime ( @xmath162 ) , and the weak bias regime in which the dominant interaction is the on - site repulsion provides us with control over the entangling interaction @xmath168 .
the boundary of the strong bias regime considered here is set by @xmath169 . a similar strong bias regime with the lowest energy singlet @xmath170 exists for @xmath171 , where @xmath172 , but we do not consider it here in detail . in both of these regimes ,
a wide energy gap @xmath173 to the second doubly occupied state allows us to neglect that state .
this approximation reduces the dimensions of the hamiltonian blocks @xmath174 , eq . ( [ eq : hblocks ] ) , and allows for a perturbative solution . since the only available dqd states in the strong bias regime are the triplet @xmath65 and two singlets , @xmath62 and @xmath63 , the @xmath175 block of eq .
( [ eq : hblocks ] ) is one - dimensional , @xmath176 and @xmath177 are two - dimensional , and @xmath178 is four - dimensional . for the present discussion of the strong bias regime ,
we choose the zero of the energy scale at @xmath179 , setting the expectation value of the energy of four singly occupied qds with the dqds in the electron singlet states to zero , @xmath180 . using the expressions for the hamiltonian matrix elements given in the appendix a , we find the matrices of the @xmath24 blocks ( @xmath25 ) .
the energy of the @xmath56 state is then @xmath181 the two - dimensional blocks @xmath177 and @xmath176 are related by the symmetry under exchange of the double quantum dots @xmath182 and in the bases @xmath183 , and @xmath184 , @xmath185 , have the identical matrix form @xmath186 where @xmath187 is the renormalized hopping matrix element and @xmath188 is the electrostatic energy cost of doubly occupying the @xmath90 state in the presence of the triplet dqd . with our choice of the zero of the energy scale ,
the ground state energies of @xmath176 and @xmath177 are @xmath189 from the energies @xmath190 and @xmath191 , we extract the isotropic exchange part of the low - energy four - spin hamiltonian eq .
( [ eq : hstrongbias ] ) as @xmath192 the resulting exchange interaction strength is @xmath193 comparing this result with the case of an unbiased isolated double quantum dot , eq .
( [ eq : jbld ] ) , we see that the effect of the strong bias @xmath18 and the presence of another dqd behind the impenetrable barrier is the change of the effective on - site repulsion to the value @xmath194 and a reduction of the effective tunneling matrix element because of the large gap to the excited doubly occupied state . as a consequence of this gap ,
the isotropic exchange in the limit of noninteracting dqds and weak tunneling is @xmath195 , with the hopping contribution reduced to a half of the result expected from the standard hubbard model in the unbiased case , @xmath196 @xcite .
the four - dimensional block @xmath178 in the basis @xmath197 is @xmath198 where @xmath199 is the tunneling matrix element renormalized by the spectator dqd in the doubly occupied state , and @xmath200 accounts for the repulsion energy of four electrons in the @xmath90 orbitals and the bias @xmath18 , see appendix a. due to the symmetry with respect to exchange of the dqds , @xmath201 , the antisymmetric state @xmath202 decouples from the other , symmetric , states . in the limit of large and positive bias , @xmath203 , all the tunnelling and correlated hopping terms in the hamiltonian @xmath178
can be taken to be small .
the unperturbed hamiltonian is then diagonal and the ground state energy is @xmath204 .
this situation is relevant , because all the small terms are proportional to the overlap @xmath26 of the localized states in the quantum dots , which is small for weakly tunnel - coupled qds , and we can reach this regime by applying external voltage to make @xmath205 large enough . operating the system in the strong bias regime causes a qualitative change to the effective low - energy hamiltonian by turning on the entanglement generating term @xmath168 in eq .
( [ eq : hstrongbias ] ) , @xmath206 for weak bias and in the absence of tunneling , the entanglement generating @xmath168 term is zero , as can be checked from the energies of the states @xmath207 , @xmath208 , and @xmath209 , given in appendix a. this is not true in the case of strong bias , where the entangling interaction of the strength @xmath210 is present even if the tunneling terms are zero . in the strong bias regime ,
the conditions for @xmath211 are @xmath212 , @xmath128 , and @xmath213 . while the first two conditions are satisfied when there is no tunneling , the third is independent of the tunneling .
it is only satisfied in the limit of long distance between dqds , @xmath214 , see fig .
[ fig:4dots ] .
the tunneling causes a second - order correction to @xmath215 , @xmath216 and the corresponding correction to @xmath168 @xcite .
we have calculated the matrix elements of the coulomb interaction using the basis of single - electron wannier states obtained by orthogonalizing the fock - darwin ground states centered at the quantum dots positions , following @xcite .
the resulting matrix elements can all be expressed in terms of the distances between the quantum dots , and the tunneling matrix element @xmath17 between qd in dqd .
these results are summarized in appendix a. together with eqs .
( [ eq : hstrongbias ] ) , ( [ eq : jstrong ] ) , and ( [ eq : ess2 ] ) , they provide a model of the low - energy hamiltonian of a pair of qubits realized on a ddqd in the strong bias regime .
this model can describe a two - qubit quantum gate realized by adiabatically switching the value of the control parameter @xmath18 so that the qubit goes from the weak bias regime to the strong bias regime and back . in summary , the interaction of the dqds causes a change in the parameters of the extended hubbard model coupling strength , eq .
( [ eq : jbld ] ) , so that the energies and hopping matrix elements on one dqd depend on the state of the other .
also , the processes in which the hopping of electrons on the two dqds is correlated and mediated by the direct coulomb interaction become possible , see fig .
[ fig : correlation ] .
the coupling between the dqds causes an effective spin interaction that deviates from the form of exchanged - coupled qubits , adding the entangling term @xmath168 to the eq .
( [ eq : hstrongbias ] ) .
this deviation creates the entanglement between the two qubits .
the generation of entanglement can be efficiently controlled by changing the bias @xmath18 .
the study of a double double quantum dot ( ddqd ) system in the strong bias regime presented in sec .
iii allows for a simple perturbative solution and offers an insight into the mechanism of entanglement generation .
however , it lacks sufficient predictive power for a general analysis of a realistic two - qubit quantum gate : when switching on and off the entangling interaction , a continuous voltage pulse is applied , and the system undergoes a smooth transition from the strong bias regime to the unbiased ( or merely biased ) regime and vice versa . during this transition ,
the system has to pass through an intermediate weak - bias regime where the perturbative expansion eq .
( [ eq : ess2 ] ) breaks down . in this section , we calculate the full hund - mulliken ( hm ) hamiltonian of the four quantum dots , including both @xmath63 and @xmath64 states .
this calculation allows us to predict the quantum gate generated by an arbitrarily shaped adiabatic pulse of the control parameters @xmath17 and @xmath18 .
the main difference in the system s description is that now we take into account both doubly occupied states @xmath63 and @xmath64 in each dqd .
therefore , we are working in the entire hilbert space of the hm approximation , and the strong bias requirement is not important .
now , @xmath175 is one - dimensional , @xmath176 and @xmath177 are three - dimensional , and @xmath178 is nine - dimensional . following the discussion of sec .
iii , the effective low - energy spin hamiltonian @xmath217 , eq .
( [ eq : hstrongbias ] ) , is determined by the energies @xmath50 , where @xmath25 , of the lowest energy states of a given spin configuration . due to the @xmath201 symmetry , @xmath217 is the sum of the isotropic exchange terms and the entangling term .
we proceed by calculating the matrix elements of the hamiltonian as a function of the tunneling matrix element @xmath17 and the bias @xmath18 .
the results of this calculation are given in appendix a. numerical diagonalization of the resulting hamiltonian gives the energies @xmath50 , for each of the blocks @xmath24 , where @xmath25 . finally , we extract the effective low - energy hamiltonian parameters @xmath35 and @xmath168 using eq .
( [ eq : jspectrum ] ) and eq .
( [ eq : ee ] ) .
the dependence of the isotropic exchange coupling on the bias @xmath218 is illustrated in fig .
[ fig : jofepsilon ] . in the zero - tunneling limit
, we can identify three regions of qualitatively different behavior of @xmath218 . for strong and negative bias , @xmath219 ,
corresponding to the @xmath165 lowest energy singlet state , the isotropic exchange coupling is decreasing linearly with the bias . in the intermediate region @xmath220 the exchange coupling is absent . for strong and positive bias @xmath221 ,
the exchange coupling grows linearly with @xmath18 .
the asymmetric placement of the @xmath222 plateau is a consequence of the different repulsion energies of the electrons in the inner and outer qds .
as the tunneling is turned on , the isotropic exchange couplings becomes larger due to the mixing of the doubly occupied states in the plateau region .
for zero magnetic field , the coupling @xmath35 is positive . in a finite field
there is a region with negative @xmath35 , consistent with the analysis of @xcite and the experimental findings of @xcite . as a function of the bias @xmath18 . in the regions of strong positive and negative bias ,
the exchange coupling is approximately linear @xmath223 . in the intermediate region ,
the exchange is zero in the zero tunneling limit and becomes nonzero as the tunneling is turned on .
the coupling @xmath35 is always positive in the absence of a magnetic field .
the external magnetic field drives @xmath35 to negative values in a relatively wide range of values of the tunneling matrix element and bias .
the confinement energy of the quantum dots is chosen to be @xmath224 , which corresponds to a quantum dot bohr radius @xmath225 in gaas .
the distances between the dots are chosen to be @xmath226 and @xmath227 .
[ fig : jofepsilon],width=321 ] a plot of the entanglement generating interaction @xmath168 is given in fig .
[ fig : general_bias ] . as a function of bias .
the plots correspond to different values of the tunneling matrix elements @xmath17 within the double quantum dots in the absence of a magnetic field and in an external magnetic field of @xmath228 .
the @xmath152 plot indicates the regions of different lowest energy singlets and the positions of crossings .
the strength of the entangling interaction @xmath229 can be changed significantly by tuning the bias @xmath18 at a fixed tunneling matrix element @xmath17 .
parameters used in this plot are the same as in fig .
[ fig : jofepsilon ] .
[ fig : general_bias],width=321 ] the zero - tunneling value of @xmath168 shows a structure determined by the coulomb energies of the basis states eqs .
( [ eq : states])([eq : statesend ] ) . in a wide plateau of small bias
the entangling interaction vanishes , because all of the lowest - energy states of definite spin are products of @xmath153 and @xmath65 .
since the direct exchange interaction @xmath230 is zero in the absence of tunneling , those two states are equal in energy .
when the bias overcomes the on - site repulsion , the lowest energy states of @xmath178 , @xmath176 , and @xmath177 change .
the degenerate lowest energy states of @xmath178 are either @xmath231 and @xmath232 in the region of large bias on the right of the plateau , or @xmath233 and @xmath234 in the region of smaller bias to the left of the plateau .
simultaneously , the analogous states with @xmath62 replaced by @xmath65 become the lowest energy states in @xmath176 and @xmath177 . in these two regions
@xmath168 is a linear function of @xmath18 , @xmath235 on the left and @xmath236 on the right of the plateau .
when the absolute value of the bias is even higher , the lowest energy state in @xmath178 is @xmath163 for a very strong and positive bias and @xmath165 for a very strong and negative bias .
these regions are characterized by an @xmath18-independent @xmath237 for large @xmath238 .
the values @xmath159 for the bias @xmath18 at which the changes in zero - tunneling lowest energy states occur depend on the geometry of the device , described by the distances @xmath137 and @xmath16 ( fig .
[ fig:4dots ] ) and the quantization energy @xmath239 , and correspond to the changes in behavior of the exchange coupling strength @xmath35 .
the zero - tunneling case shows a desirable feature in that @xmath168 , the quantity that determines the entanglement between the qubits , can be switched on and off by tuning @xmath18 .
however , the regions of different @xmath168 can not be reached by adiabatic pulses in the @xmath240 limit .
turning on the tunneling @xmath17 between the qds will introduce transitions between previously disconnected regions , and the adiabatic gates become possible .
the simple @xmath152 picture of the entanglement generated by a difference in coulomb energies is perturbed by the transitions .
it is no longer possible to turn off @xmath168 throughout the plateau region by a change in @xmath18 alone . in the plateau region
, @xmath168 is generically nonzero , but small .
therefore , in order to turn off the entangling interaction when @xmath17 is kept constant , it is desirable to keep @xmath17 small , and to tune @xmath18 to a value where @xmath211 .
for a quantum gate applied by the time - dependent hamiltonian eq .
( [ eq : hblocks ] ) , with the parameters @xmath17 and @xmath18 changing adiabatically on the time scale set by the energy gap between the states within the blocks @xmath24 , the applied gate is determined by the splittings between the lowest lying states in each of the subspaces of the definite spin . if the energies of the lowest energy states in singlet - singlet , singlet - triplet , triplet - singlet and triplet - triplet subspaces are @xmath241 , @xmath242 , and @xmath243 respectively , the gate applied by an adiabatic pulse starting at the time @xmath244 and finishing at @xmath245 will be @xmath246 , with the phases @xmath247 with the ability to turn the entangling interaction on and off and perform single - qubit gates , it is possible to perform a cnot gate on a pair of qubits encoded into spin states of dqd .
we consider a quantum gate implemented by first adiabatically turning on the entangling interaction for a period @xmath248 , and then again adiabatically switching to the hamiltonian with the entangling interaction off for the time interval @xmath249 .
the lowest energy states in each of the @xmath250 , @xmath251 , @xmath252 , and @xmath253 subspace will acquire a phase dependent on the control parameters @xmath18 and @xmath17 and the pulse durations . in the _ on _ state ,
the hamiltonian that describes the ground states in all the spin subspaces is , up to a constant , @xmath254 , where @xmath168 is the strength of the entangling interaction in the _ on _ regime , and @xmath255 is the corresponding exchange coupling .
after the ddqd was in the on state for the time @xmath256 , the applied gate is @xmath257 similarly , during the subsequent period of duration @xmath249 when the entangling interaction is set to zero , the applied gate is @xmath258 where @xmath259 in analogy with the _ on _ regime .
the resulting gate is @xmath260 where @xmath261 is the integrated strength of the exchange coupling in dqd , and @xmath262 is the integrated strength of the entangling interaction .
the cphase gate , which is equivalent to cnot up to single qubit rotations , is obtained when the gate parameters satisfy @xmath263 and @xmath264 , for an odd integer @xmath77 and an arbitrary integer @xmath76 . in order to complete a cnot
, we follow a pulse of on - state hamiltonian of the duration @xmath265 by a pulse of the off - state hamiltonian with of the duration @xmath266 .
the resulting gate is @xmath267 , for odd @xmath76 and @xmath268 , which is equal to @xmath269 with the @xmath270 gate applied to both qubits before and after @xmath271 . for any integer @xmath76 , @xmath272 where @xmath273 . in order to complete the cnot
, we apply the one - qubit haddamard gates @xmath274 to the target qubit both before and after the entangling gate @xmath271 .
the entire construction can be represented as @xmath275 note that the cnot construction necessarily involves the single qubit rotations about pseudospin axes different from @xmath13 .
such operations can be performed using the asymmetric bias within a dqd that encodes the qubit in an inhomogeneous external magnetic field @xcite .
the entangling part of a cnot gate _ _ c__an be performed by pulsing the bias @xmath18 only , and keeping the tunneling @xmath17 constant . therefore , control over the bias @xmath18 and the availability of an inhomogeneous magnetic field are sufficient for the universal quantum computing with two - spin qubits .
we have analyzed two - qubit gates in a pair of qubits , each encoded into singlet and triplet states of a dqd , and coupled by coulomb repulsion .
a two - qubit @xmath276 gate , which together with the single qubit rotations forms a universal set of quantum gates , can be performed by tuning the bias of the inner dots with respect to the outer ones .
we identify the entangling interaction strength @xmath277 as a quantity that has to be controlled in order to implement a @xmath276 with the aid of single qubit rotations .
the dependence of @xmath229 on the externally controllable bias @xmath18 and the tunneling matrix element @xmath17 shows that it can in principle be turned on and off by changing @xmath18 alone , if sufficiently low values of @xmath17 are available .
the largest change in @xmath229 comes from tuning of the system through the resonance between singly occupied state and doubly occupied state on a dqd . at the side of the resonance with a singly occupied ground state , and far from the resonance , the entangling interaction @xmath168
is caused by inter - dqd correlation and is small . on the other side of the resonance , with a doubly occupied dqd ground state ,
the entangling interaction is caused by the direct coulomb repulsion and it is much stronger .
two - qubit gates necessary for a universal set of gates can be performed by switching between the strong and weak entanglement generation regimes using voltage pulses .
we thank m. trif and d. klauser for discussions .
we acknowledge funding from the swiss national science foundation ( snf ) and through nccr nanoscience .
the full hund - mulliken hamiltonian is block - diagonal due to the symmetry of the interactions with respect to arbitrary rotations in spin space . in reality , this symmetry is broken by the weak spin - orbit coupling interaction that we have neglected .
the blocks are the one - dimensional @xmath175 , the two three - dimensional @xmath177 and @xmath176 , and the nine - dimensional @xmath178 , where @xmath27 stands for a triplet and @xmath26 for a singlet state on a dqd . in this appendix
, we present the matrices of these blocks as functions of the system geometry and the control parameters .
there is only one @xmath253 state and its energy is @xmath279 the three - dimensional blocks @xmath177 and @xmath176 are related by the symmetry operation of exchanging the dqd and if we choose the basis @xmath280 for the @xmath252 and @xmath281 for the @xmath251 subspace , they can both be represented by the matrix @xmath282 the nine - dimensional block of singlet states , in the direct product basis composed out of the two - electron states eq .
( [ eq : states ] ) is we do not antisymmetrize with respect to the permutations of electrons that belong to different quantum dots and have non - overlapping orbital wave functions .
the matrix elements of the hamiltonian that describe the coulomb interaction within a dqd ( intra - dqd terms ) @xmath58 , @xmath17 , @xmath270 , @xmath284 and @xmath285 were analyzed in @xcite .
the inter - dqd elements depend on the following matrix elements of the coulomb interaction between the product states of the @xmath286 electrons localized in the qubit @xmath287 and the quantum dot @xmath288 , and having a spin @xmath289 , @xmath290 in zero magnetic field , we find that @xmath128 . to represent the matrix elements in terms of the system parameters , the single qd quantization energy @xmath295 ,
tunneling matrix element within an isolated dqd @xmath17 , the bias @xmath18 and the interdot distances @xmath296 and @xmath297 , we have to adopt a model for the binding potential of a dqd and the orbitals of hund - mulliken approximation .
we assume that the qd orbitals are wannier functions obtained by orthogonalization of the fock - darwin ground states centered at the positions of the qds within a dqd , @xmath298 and @xmath299 .
the wannier orbitals are of the generic form @xmath300 where @xmath301 is the fock - darwin ground state on the dot belonging to the qubit @xmath302 and the inner@xmath303 or outer@xmath304 qd , eq .
( [ eq : fd0 ] ) .
the wannier orbitals are determined by the overlap of these wave functions , @xmath305 , through the mixing @xmath306 and normalization constant @xmath307 .
@xmath310 , % \\
% \label{eq : uf } u_f&=&cn^4\left [ \vphantom{\frac{d_1}{2 } } f(d_2+d_1,0 ) + 2g^2\left(1+s^2\right)f(d_2,0 ) + g^4f(d_2-d_1,0 ) + 2s^2g^2f(d_2,d_1 ) - \right . \\
\nonumber & & \left.4gs\left(f(d_2+\frac{d_1}{2},\frac{d_1}{2 } ) + g^2f(d_2-\frac{d_1}{2},\frac{d_1}{2})\right ) \vphantom{\frac{d_1}{2 } } \right ] , % \\
% \label{eq : um } u_m&=&cn^4\left [ \vphantom{\frac{d_1}{2 } } ( 1+g^4)f(d_2,0 ) + g^2\left(f(d_1+d_2,0 ) + f(d_1-d_2,0 ) + 2s^2\left(f(d_2,0 ) + f(d_2,d_1)\right)\right ) - \right . \\
\nonumber & & \left .
2gs\left(1+g^2\right ) \left(f(d_2+\frac{d_1}{2},\frac{d_1}{2 } ) + f(d_2-\frac{d_1}{2},\frac{d_1}{2})\right ) \vphantom{\frac{d_1}{2 } } \right ] , % \\
% \label{eq : to } t_o&=&cn^4\left [ \vphantom{\frac{d_1}{2 } } s\left(\left(1 + 3g^2\right)f(d_2+\frac{d_1}{2},\frac{d_1}{2 } ) + \left(g^4 + 3g^2\right)f(d_2-\frac{d_1}{2},\frac{d_1}{2})\right ) - \right . \\
\nonumber & & \left .
\left(g+g^3\right)\left(\left(1+s^2\right)f(d_2,0 ) + s^2f(d_2,d_1)\right ) - gf(d_2+d_1,0 ) - g^3f(d_2-d_1,0 ) \vphantom{\frac{d_1}{2 } } \right ] , % \\
% \label{eq : ti } t_i&=&cn^4\left [ \vphantom{\frac{d_1}{2 } } s\left(\left(1 + 3g^2\right)f(d_2-\frac{d_1}{2},\frac{d_1}{2 } ) + \left(g^4 + 3g^2\right)f(d_2+\frac{d_1}{2},\frac{d_1}{2})\right ) - \right . \\
\nonumber & & \left.\left(g+g^3\right)\left(\left(1+s^2\right)f(d_2,0 ) + s^2f(d_2,d_1)\right ) - gf(d_2-d_1,0 ) - g^3f(d_2+d_1,0 ) \vphantom{\frac{d_1}{2 } } \right ] , % \\
% \label{eq : xs } x_s&=&cn^4\left [ \vphantom{\frac{d_1}{2 } } \left(s^2 + 2g^2+g^4s^2\right)f(d_2,0 ) + g^2\left(f(d_1+d_2,0 ) + f(d_1-d_2,0 ) + 2s^2f(d_2,d_1)\right ) - \right . \\
\nonumber & & \left.2s\left(g+g^3\right)\left(f(d_2+\frac{d_1}{2},\frac{d_1}{2 } ) + f(d_2-\frac{d_1}{2},\frac{d_1}{2})\right ) \vphantom{\frac{d_1}{2 } } \right ] , % \\
% \label{eq : xd } x_d&=&cn^4\left [ \vphantom{\frac{d_1}{2 } } s^2\left(1+g^4\right)f(d_2,d_1 ) + g^2\left(f(d_1+d_2,0 ) + f(d_2-d_1,0 ) + 2\left(1+s^2\right)f(d_2,0)\right ) - \right . \\
\nonumber & & \left.2s\left(g+g^3\right)\left(f(d_2+\frac{d_1}{2},\frac{d_1}{2 } ) + f(d_2-\frac{d_1}{2},\frac{d_1}{2})\right ) \vphantom{\frac{d_1}{2 } } \right],\end{aligned}\ ] ] in terms of the overlaps of the harmonic oscillator wave functions @xmath26 , the mixing factor @xmath10 , and the function @xmath311 where @xmath312 .
we use the contraction factor @xmath148 to measure the magnetic field strength .
the overall strength of the coulomb interaction is set by @xmath313 , where @xmath314 is the electron charge , @xmath315 is the dielectric constant , and @xmath295 is the single isolated qd quantization energy @xcite .
to model the dependence of the matrix elements on externally controllable tunneling matrix element @xmath17 , we use the connection between the tunneling and the overlap @xmath316 that holds for the quartic double well , eq . ( [ eq : todb ] ) and
assume that it holds throughout the gate operation .
we have done a similar calculation of the second - order correction to the @xmath168 in the weak bias regime , valid in the region where @xmath317 .
the resulting correction to the @xmath211 zeroth - order result contains terms of the order @xmath318 and @xmath319 .
it is caused by the correlated hopping processes , and its size is @xmath320 . | we study the two - qubit controlled - not gate operating on qubits encoded in the spin state of a pair of electrons in a double quantum dot .
we assume that the electrons can tunnel between the two quantum dots encoding a single qubit , while tunneling between the quantum dots that belong to different qubits is forbidden .
therefore , the two qubits interact exclusively through the direct coulomb repulsion of the electrons .
we find that entangling two - qubit gates can be performed by the electrical biasing of quantum dots and/or tuning of the tunneling matrix elements between the quantum dots within the qubits .
the entangling interaction can be controlled by tuning the bias through the resonance between the singly - occupied and doubly - occupied singlet ground states of a double quantum dot . |
t tauri stars ( tts ) are pre - main sequence low mass stars frequently described as young analogs of our solar system .
surrounded by circumstellar disks of gas and dust , these systems mark the earliest stages of planet - formation , a process now understood to be common in our galaxy @xcite . due to their close proximity to the parent star
, intense ultraviolet and x - ray radiation fields greatly impact the evolution of the protoplanetary disk .
the far - ultraviolet ( fuv , @xmath4ev ) field is of particularly broad interest as it provides thermodynamical gas heating through the photoelectric effect @xcite , and affects disk chemistry directly through the photo - dissociation and -ionization of molecules and atoms ( * ? ? ? * and references therein ) .
fuv photodesorption of ices contributes to the transport of material between solid and gas phases , whilst enabling complex molecule formation on grains surfaces @xcite . alongside x - rays and cosmic - rays ,
the fuv field is also involved in the ionization balance of the disk ; of central importance in dynamical processes such as the magneto - rotational instability that is believed to play a role in accretion @xcite .
eventually the dispersal of disk gas may be driven by uv photo - evaporation @xcite , although resolving the contributions made by the separate fuv , euv and x - ray components is still a matter of debate @xcite one feature of the fuv spectra in t tauri systems which separates the accretors from the non - accretors , is a strong , broad @xmath0 line ( accounting for as much as 80% of the total fuv luminosity ) . in nearby systems such as tw hydra the @xmath0 line can be detected directly @xcite , whilst in obscured systems its presence may be inferred through the excitation of @xmath5 molecules @xcite .
it was soon realized that the concentration of uv photons in a relatively narrow wavelength range ( @xmath6 ) would have important implications for wavelength - dependent chemical processes , specifically the photodissociation of molecules @xcite .
for example , the molecules hcn , oh , h@xmath7o and h@xmath7co can all be photodissociated by a @xmath0 photon , unlike the closely related species cn , oh@xmath8 , h@xmath7o@xmath8 and hco@xmath8 ( for a more complete list , see * ? ? ? * ) .
abundance ratios frequently used as chemical diagnostics , such as @xmath9 , will be changed ( in this case increased ) by @xmath0-dominated photodissociation .
@xcite have shown quantitatively that the inclusion of differential photodissociation by @xmath0 can some instances help bring abundance ratios into better agreement with observation @xcite . while the the presence of @xmath0 decreases the abundance of hcn , nh@xmath10 , and ch@xmath11 by an order of magnitude or more , other species such as co@xmath7 and so are enhanced .
so actually has a significant cross - section at 1216@xmath12 , however so does so@xmath7 , and it is the photodissociation of so@xmath7 which partially replenishes the so abundance . in this instance , the presence of @xmath0 actually increases the so abundance .
similarly , the abundances of h@xmath7o and oh in the @xcite models increased with the introduction of @xmath0 , even though both h@xmath7o and oh have a significant photodissociation cross - sections at 1216@xmath12 . in this case the gas phase abundances are more than adequately replenished by efficient @xmath0 photodesorption from grain surfaces .
further chemical modeling is required before the implications of @xmath0-dominated photochemistry are fully understood . despite the interesting chemical implications of a @xmath0 dominated fuv field , the problem of precisely how it penetrates
the disk has received little attention in the literature . in this paper
we attempt to shed light upon its basic transport .
although frequently flared , protoplanetary disks remain relatively flat objects with vertical scaleheights that are small compared to the radial extent of the disk @xcite . as a result
, the majority of the mass ( residing near to the disk midplane ) is concealed from the parent star .
visual optical depths at the midplane , measured along lines directly from the star , typically exceed @xmath13 , implying that essentially no stellar photons can penetrate these parts directly . instead , radiation must be scattered downwards from the surface of the disk @xcite .
scattering is therefore of central importance when discussing the transport of stellar radiation in protoplanetary disks .
the penetration and absorption of x - ray and fuv radiation determines much of the thermodynamical stratification of the upper disk ( e.g. * ? ? ? * and references therein ) . for simplicity
we can identify three layers ; the uppermost layer being a photodissociation layer illuminated directly by the intense , unattenuated stellar radiation field - this layer is dominated by atoms , since the photodissociation timescale of molecules is very short .
as unscattered photons penetrate into the disk they eventually reach an irradiation surface , defined to be where the optical depth integrated from the star is of order unity , @xmath14 @xcite .
it is here that , in an average sense , photons undergo their first interactions with the disk . in the case of interactions with dust grains , this results in attenuation through both absorption and scattering ; the latter leading to the emergence of a diffuse component to the radiation field .
since the radiation field decreases with depth , while the local mass density increases , we eventually enter an intermediate warm molecular layer , where @xmath5 , co and a host of other gas - phase molecules may exist in a dynamic equilibrium with the photodissociating fuv field .
it is in this warm molecular layer that the photochemical processes are most important . despite its attenuation ,
the radiation field is sufficient to maintain grain temperatures above @xmath15k , thus preventing widespread freeze - out of molecules onto grain surfaces . at even greater depths
we enter the optically thick , dense midplane region containing @xmath16 of the disk mass . here
the stellar fuv field is highly attenuated and consists entirely of scattered photons . with the exception of the innermost few au ,
the temperatures may be low enough ( @xmath17k ) to permit the freeze - out of several molecular species @xcite .
the transport of fuv photons is further complicated by the gradual growth and settling of dust grains as the disk evolves towards planetesimal formation @xcite .
not only does grain - growth affect the optical properties of individual grains , but the radial and vertical drift of solids relative to gas causes a segregation of the grain - size population @xcite .
the majority of the grain opacity at fuv wavelengths is attributable to relatively small grains with radii @xmath18 m , which subsequently reemit absorbed energy at infrared wavelengths , making it possible to estimate their abundance @xcite .
observationally , the abundance of small grains ( per h nucleus ) relative to that found in the ism , @xmath19 , typically takes a value much less than one ; for example , in the taurus star - forming region the median value is @xmath20 @xcite .
our basic picture of fuv transport is somewhat incomplete once we include @xmath0 photons . in view of the large cross - section due to resonant scattering by h atoms , @xmath21
, it seems entirely plausible that @xmath0 photons experience their first interactions not with dust grains but with h atoms residing in the uppermost photodissociation layer of the disk ( hereafter ` h - layer ' ) .
the irradiation surface for @xmath0 photons is therefore defined as the location where @xmath22 .
the precise location depends quite sensitively on the width of the stellar @xmath0 line profile , although the @xmath0 photons that propagate freely to the disk surface are generally those in the wings of the scattering line profile presented by the disk @xcite .
even hundreds of doppler widths out in the lorentz wings of the scattering line profile the cross - section may still be significantly larger than that due to the ( possibly settled ) dust population .
figure [ fig : voigt ] shows the @xmath0 line from tw hya , plotted next to an estimate of the scattering line profile due to warm h atoms in the disk . assuming the photodissociation layer supports h columns of @xmath23@xmath24 ( integrated along a line from the star )
, it seems probable that the @xmath0 field will first be scattered by this atomic h - layer . the optical depth of dust in this layer
will be smaller by a factor of approximately @xmath25 .
in stark contrast to @xmath0 transport , fuv - continuum photons will stream unimpeded through the h - layer , eventually striking the disk at the conventional irradiation surface defined by dust . here
they will begin to scatter , however scattering by typical interstellar dust populations is considered to be somewhat forward - throwing @xcite .
multiple scatterings are therefore required before an appreciably diffuse component is generated , and accompanying this scattering will be a degree of pure absorption . in this paper
we explore numerically the basic phenomenology of resonantly scattered @xmath0 in disk models .
an emphasis is placed on contrasting the transport of @xmath0 with that of the fuv - continuum field in the vicinity of the @xmath0 line .
we proceed with section [ sec : disks ] where we discuss the properties of our disk models .
radiation transfer of @xmath0 and fuv - continuum photons is dealt with in section [ sec : radiative - transfer ] .
this includes an approximate - but largely self - consistent -computation of the h distribution necessary for @xmath0 transport .
results are presented in section [ sec : results ] .
the paper concludes with a discussion and summary .
we base our disk structures on a suite of three disk models from @xcite .
the disks are differentiated by the depletion of small grains in their upper layers , @xmath26 .
physically , the parameter @xmath19 represents the ratio of dust concentration relative to that found in the interstellar medium . although the models include the effects of dust depletion on the hydrostatic disk structure , they lack a separate thermodynamic treatment of the gas , which becomes thermally decoupled at the low densities found in the disk atmosphere . while the dalessio et al .
disks suffice for illustrating the essential features of @xmath0 and fuv - continuum transport , a more self - consistent treatment will eventually require the inclusion of detailed coupling between the radiative transfer , thermodynamics and disk structure .
several disk models exist that already include the thermal decoupling of gas and dust that generally leads to larger disk scaleheights ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
we assign a nominal fuv luminosity of @xmath27 to the central star and a kinetic temperature of @xmath28k to the gas in the upper layers of the disk . in view of these modifications
we refer to the models as ` modified dalessio disks ' .
a cross - section of the total hydrogen density , @xmath29 , in one of the disk models is shown in figure [ fig : nh_map ] .
the grain - size distribution used in the original dalessio calculations follows @xcite , @xmath30 , where @xmath31 is the grain radius .
in fact there are two populations of grains ( ` large ' and ` small ' ) in the dalessio models , however the large grains are confined to the disk midplane and play no role in what follows . for the ` small ' grain population
the grain - size limits are @xmath32 m and @xmath33 m ( double check this ) .
these limits are consistent with those proposed by @xcite for interstellar grains .
relevant optical properties of the ` small ' grain population are shown in figure [ fig : dust_properties ] . in this paper
we are primarily concerned with fuv - continuum wavelengths in the vicinity of @xmath34 . at this wavelength
the ` small ' grain population exhibits an appreciably forward - throwing phase function ( asymmetry factor of @xmath35 ) .
the widely used henyey - greenstein phase functions @xcite evaluated for @xmath36 and @xmath37 are shown in figure [ fig : hg ] . the computational domain used for the subsequent radiative transfer simulations comprises the modified dalessio disks immersed in a background mesh .
the modified dalessio disks are spatially 1 + 1d datasets ; the purpose of the background grid of nodes is to transform these models into true 2d distributions .
the resulting discretization consists of a list of nodes , each of which has associated with it a location as well as relevant physical quantities ( e.g. densities ) which are interpolated from the original dalessio models .
the background mesh supports unstructured distributions of nodes in order to accommodate spatially irregular datasets such as the dalessio models . to the extent possible , the initial background grid exhibits grid refinement such that its spatial resolution matches that of the dalessio data .
the nodes ( constrained to lie in the @xmath38 plane in a cylindrical coordinate system ) ultimately form the vertices of triangular cells .
figures [ fig : grid_adaptive ] and [ fig : grid_inner ] give the reader an impression of the spatial distribution of nodes and how they result in a tessellation of triangular cells . the triangular cells are a result of the connectivity obtained by performing a delaunay tessellation on the node list . in 2d
the triangle is the simplex : the simplest shape that can tessellate the volume @xcite . in 3d
the simplex is the tetrahedron .
there are many alternative ways to connect the nodes that give rise to a sensible tessellation of triangular cells , however the delaunay tessellation has unique properties that arguably yield the optimal connectivity .
first , the connections link those nodes that compete for space .
this is most easily understood by considering the corresponding voronoi diagram ( the dual of the delaunay tessellation ) .
second , the resulting triangles maximize the minimum internal angle , thus reducing the frequency of skinny triangles . while the meshes in figs .
[ fig : grid_adaptive ] and [ fig : grid_inner ] ostensibly appear regular , they are treated as fundamentally unstructured .
had the background mesh been cast randomly ( e.g. according to a poisson point process ) it would be possible to shed all traces of the underlying coordinate system in the spatial discretization @xcite . for these reasons the delaunay tessellation is becoming an attractive choice for generating high - quality , unstructured discretizations of spatial objects .
very few radiative transfer problems of astrophysical interest can be solved analytically without resorting to significant simplifications @xcite .
the transport of radiation in protoplanetary disks is no exception . in view of the geometrical complexity and assortment of opacity sources , we opt to solve the transport problem using the highly versatile monte carlo simulation method .
numerical details of our monte carlo radiative transfer code are described in the appendices . in this section
we restrict the discussion to aspects which are relevant to the physical context .
the transport of fuv - continuum photons is assumed to be governed solely by interactions with dust grains . at extreme levels of dust settling ( @xmath39 )
it is possible that rayleigh scattering by @xmath5 molecules becomes relatively important .
anisotropic scattering by grains is treated using the one - parameter henyey - greenstein phase function , the angular scattering probability distribution of which is given by , @xmath40^{3/2}}.\label{eq : hg}\ ] ] where @xmath41 $ ] is the anisotropy parameter .
isotropic scattering corresponds to @xmath42 , forwards scattering to @xmath43 , and backwards scattering to @xmath44 ( figure [ fig : hg ] ) .
the single - scattering albedo @xmath45 indicates that these grains have a slight preference towards absorption rather than scattering . from @xmath46 and @xmath47
we can conclude that multiple scatterings will be required in order to generate a diffuse field that can penetrate the disk vertically .
accompanying these scatterings will be significant absorption .
the transport of @xmath0 is treated like fuv - continuum but with the addition of resonant scattering .
accordingly , it is necessary to first determine the distribution of atomic hydrogen .
the formation of @xmath5 in dense , dusty environments usually occurs on the surfaces of dust grains @xcite .
the destruction of @xmath5 occurs with @xmath49% probability following an absorption of a fuv photon into the lyman - werner band system ( @xmath50 , * ? ? ?
since the photodissociation of @xmath5 occurs through line absorptions , these transitions become optically thick over very modest columns , @xmath51@xmath24 . beyond this point
we see a run - away formation of ` self - shielded ' @xmath5 . while any photo - destruction dominated process that competes with formation can in principle result in self - shielding @xcite , the concentration of @xmath5 opacity in discrete lines makes it readily self - shielding . balancing formation and destruction rates
gives the steady - state molecular fraction @xcite , @xmath52 where @xmath53 is the high temperature @xmath5 formation rate coefficient @xcite , and @xmath54 is the @xmath5 photodissociation rate @xmath55 $ ] . for the self - shielding function , @xmath56 $ ] ,
we adopt the closed - form expression from @xcite , although comparison simulations were also performed using the tabulated results of @xcite .
these approximations were shown to result in h - layers with comparable geometries and column densities . in the context of our monte carlo simulation
we evaluate the self - shielding function along the length of each photon packet trajectory ( e.g. * ? ? ?
the inclusion of a full treatment of @xmath5 photodissociation is currently beyond the capabilities of most multidimensional photochemical codes @xcite .
the calculation of the h/@xmath5 distribution is an iterative one , since to compute @xmath57 at a point requires us to know the degree of self - shielding due to @xmath57 everywhere else .
an initial distribution of h and h@xmath7 is required to start the iterative calculation . from this our @xmath5 will ` grow ' as self - shielding is established over the course of repeated iterations . ideally , this initial condition should be constructed to underestimate the @xmath5 density : if @xmath5 is overestimated then regions that are artificially self - shielded will be slowly photodissociated away over the course of the iterative process . in such over - shielded cases , a layer of excess h@xmath7 with a thickness corresponding approximately to the self - shielding column ( @xmath58@xmath24 )
will be photodissociated in each iteration .
if the excess @xmath5 is distributed over a large volume , the convergence towards the true h@xmath7 distribution will be very slow .
since the introduction of self - shielding always _ increases _ the h@xmath7 density , we can set a strict lower limit by computing @xmath57 using eqn [ eq : h2 ] ignoring all forms of shielding ( this includes dust attenuation , although we retain the geometrical inverse - square dilution of radiation ) .
the resulting @xmath59 is shown in the left - hand panel in figure [ fig : grid_adaptive ] . even without self - sheilding ,
atomic hydrogen is limited to the upper disk layers due solely to the increase in density that promotes h@xmath7 formation deep in the disk .
the iterative process commences with a full radiative transfer simulation at a single representative lyman - werner band wavelength ( centered at @xmath60 ) .
the accumulation of h@xmath7 column density along photon packet trajectories is recorded , and information regarding the self - shielded radiation field is deposited in each cell traversed by the photon packet . dust scattering and absorption
are both included .
however , since @xmath5 shielding proceeds so vigorously , dust shielding plays a relatively minor role in determining the final h/@xmath5 distribution .
this is especially true in dust - settled scenarios , although a rigorous consideration must establish the criteria that separates purely @xmath5 self - shielding situations from those in which @xmath5 self - shielding is initiated by dust attenuation @xcite . once approximately @xmath61 photon packets have been run , the resulting self - shielded radiation field is used to compute a new @xmath57 using eqn [ eq : h2 ] .
since the original grid is generally too coarse to properly resolve the emerging h / h@xmath7 transition we refine the grid before proceeding with a new iteration .
if the @xmath5 distribution is described on grid that is too coarse , individual cells may self - shield , causing a local overestimation of @xmath62 that may subsequently propagate through the grid .
the refinement criterion is based upon the fractional change in @xmath57 between iterations .
this ensures that the spatial resolution of the grid is increased in regions which are most sensitive to changes in @xmath57 .
regions that are either entirely unshielded or totally self - shielded ( @xmath63 ) will not receive grid refinement .
the refinement of a cell is achieved by placing a single node in the geometric center of the cell .
once new nodes have been added a delaunay tessellation is performed on the augmented node list , resulting a new set of cells .
physical quantities are interpolated onto this new grid and the next iteration cycle is begun . the criteria for a converged solution must not only require a solution that does not change significantly with further iterations , but
also that every cell in the h/@xmath5 transition must not contain a sufficiently large @xmath64 to self - shield all by itself .
for a cell to be considered part of the h/@xmath5 transition it must exhibit a @xmath57 that is greater than its unshielded value and less than its maximal possible value of @xmath65 .
the nodes and @xmath59 distributions after the first and second iterations in the @xmath66 model are shown in figure [ fig : grid_adaptive ] .
it is worth noting that the uppermost layers are entirely photodissociated so that @xmath67 throughout the iterative procedure .
it is the _ lower _ boundary of the h layer that is affected by self - shielding , and this can be seen to rise in height as the bulk of the disk gradually becomes molecular . for the modified dalessio disks we typically undertake 15 - 20 iterations .
convergence is assumed when @xmath57 changes by less than 5% everywhere in the domain .
it is important to realize that monte carlo noise will generate variance between iterations , even after many iterations .
if the monte carlo noise level is known then eqn [ eq : h2 ] can be used to estimate its effect on @xmath57 . to summarize ,
the basic steps of the iterative h@xmath7 calculation are as follows ; 1 .
compute an initial distribution ( lower limit ) of h@xmath7 due to dissociation by the unattenuated stellar field ( i.e. ignoring @xmath5 self - shielding and dust attenuation ) .
2 . propagate photon packets using monte carlo methods , following @xmath5 self - shielding and dust - absorption along trajectories .
compute new h@xmath7 distribution due to this photodissociation field .
4 . add nodes to regions where the h / h@xmath7 transition is detected .
interpolate physical properties onto the new mesh . 5 .
repeat from step 2 until solution has converged .
once the h - layer has been calculated it is possible to propagate @xmath0 photons .
the resonant scattering process is simplified somewhat by associating with it a voigt line profile , @xmath68 , and an isotropic phase function . @xcite and
@xcite discuss the treatment of resonant @xmath69 transitions in more detail , although for our purposes the simplified treatment is sufficient . as discussed previously , most @xmath0 photons from the tw hya spectrum scatter in the lorentzian wings of the voigt profile ( see figure [ fig : voigt ] ) . unlike the doppler core , which is relatively temperature sensitive , the wings of the scattering profile do not change greatly with the temperature of the gas .
this somewhat mitigates the ad hoc temperature imposed upon the gas in our modified dalessio disks .
the convolution of the gaussian ( thermal broadening ) and lorentz ( natural ) line profiles gives rise to the voigt profile , @xmath70 where @xmath71 is the oscillator strength for the the @xmath72 transition , @xmath73 is the electron charge , @xmath74 the speed of light , and @xmath75 the electron mass .
@xmath76 is the thermal doppler width , where @xmath77 is the mean thermal velocity of the scatterers and @xmath78 is the frequency at line center .
@xmath79 is the voigt profile defined by the integral , @xmath80 while the scattering is coherent in the frame of the scattering atom , an overall frequency change occurs due to the difference between the incoming and outgoing doppler shifts required to change between atom and disk frames .
this is known as partial frequency redistribution @xcite .
the fractional change in photon frequency is of order the thermal doppler width @xmath76 where @xmath77 is
the mean thermal velocity of the scatterers and @xmath78 is the frequency at line center . partial redistribution of the scattered @xmath0 photon is included in our code using microscopic monte carlo techniques based on @xcite and @xcite .
however , since the intrinsic stellar @xmath0 line shape is already very broad ( @xmath81 doppler widths , see fig .
[ fig : voigt ] ) , and the h - layer not extremely optically thick , little appreciable frequency redistribution of the @xmath0 line is expected .
as such one can largely ignore frequency redistribution in our models , and instead treat the resonant scattering as elastic . this simplification greatly expedites the calculation .
we ignore systematic doppler shifts due to the azimuthal velocity component of the disk ( @xmath82kms@xmath83 at 1au , * ? ? ?
* ) since for much of the disk this induces doppler shifts that are modest compared to the stellar @xmath0 line width ( @xmath84kms@xmath83 ) .
the vertical distribution of hydrogen in the three disk models is shown in figure [ fig : columns ] .
the molecular hydrogen column density @xmath64 shows the characteristic s - shape associated with the onset of self - shielding as we descend into the disk .
the vertical column density of atomic hydrogen clearly varies with the degree of dust depletion , @xmath85 - this is the column thickness of the h - layer that scatters @xmath0 photons .
thicker h - layers are associated with dust depleted models , due to the proportional reduction in the @xmath86 formation rate .
the precipitous drop in @xmath59 seen in figure [ fig : columns ] is indicative of the sudden onset of @xmath5 self - shielding .
the _ vertical _ optical depth ( for resonant scattering ) of the h - layer depends upon the wavelength displacement from line center ; for the @xmath0 spectrum shown in figure [ fig : voigt ] the photon - averaged optical depths are @xmath87 for @xmath88 , respectively .
although these vertical optical depths are not large , when viewed from the star the _ slant _ optical depth this h - layer is typically @xmath89 and thus _ the h - layer will intercept essentially every stellar @xmath0 photon incident upon it_. we are mostly interested in following the portion of @xmath0 emerging from the lower surface of the h - layer .
the remainder is scattered into space from the upper surface .
it is worth noting that the h - layer is not so optically thick to these line - wing photons that they become line - trapped and destroyed by dust absorption @xcite .
while these results suggest that resonant scattering of @xmath0 is indeed important , the h - layers computed lack the self - consistency required to render them entirely realistic .
figure [ fig : maps ] shows spatial maps providing a visual comparison of the @xmath0 and fuv - continuum photon densities in the @xmath66 modified dalessio disk .
in addition to the photon density @xmath90 we also show flux arrows ( direction only ) , and the anisotropy of the field @xmath91 ( the ratio of net flux to photon density ) . the highly flared h - layer and its isotropizing effect on the @xmath0 field is clearly seen in the left - hand anisotropy panel .
the net flux of transmitted @xmath0 photons emerges from the h - layer perpendicularly , providing a more direct illumination of the disk below .
vertical profiles of the @xmath92fuv - continuum photon density ratio are shown in figure [ fig : ratios ] for @xmath93 and @xmath94 , at radii @xmath95 and @xmath96au .
the same qualitative behavior of the ratio is seen at all radii , although the quantitative effect is greatest in the inner disk .
it is clear from figure [ fig : ratios ] that the h - layer is fuv - continuum dominated , whereas the molecular disk ( and therefore the vast majority of disk mass ) is @xmath0-dominated .
ultimately , the enhancement of the @xmath0 photon density ( above the intrinsic stellar value ) may exceed an order of magnitude .
the ratios eventually asymptote to a constant value deep into the disk where both fields are approaching the diffusive limit and therefore behaving similarly .
regardless of the value of @xmath19 , essentially every stellar @xmath0 photon is intercepted by the highly flared h - layer . upon the first scattering the @xmath0 field
is completely isotropized and a fraction @xmath97is transmitted through to the base h - layer .
this is clearly seen in fig .
[ fig : maps ] ( top left and bottom left panels ) . viewed from within the molecular region the @xmath0 would seem to originate in a diffuse blanket overlying the disk , analogous to the diffuse transport of sunlight on an overcast day .
this is similar to the penetrative advantage that interstellar photons ( bathing the disk isotropically ) would have ( e.g. * ? ? ?
* ) , however in our models the stellar @xmath0 field is typically orders of magnitude more intense .
once in the realm of the molecular disk , @xmath0 transport will be controlled primarily by dust .
fuv - continuum photons pass unimpeded through the h - layer until they reach the dust irradiation surface . until this point the field only experiences geometric inverse - square law diminution - in contrast to the @xmath0 , which has been additionally processed by the h - layer . in this regime
the @xmath0/fuv - continuum photon density ratio favors the unimpeded fuv - continuum photons .
the fuv - continuum eventually impinges upon the dust irradiation surface , doing so at very small angles ( @xmath98 degrees ) .
the relative inefficacy of scattering by dust grains is exacerbated by accompanying absorption .
it is straightforward to show with 1d planar radiative transfer models that the energy density transmitted downwards from this dust irradiation surface is typically of order @xmath99 of that incident upon it @xcite .
consequently , the @xmath0/fuv - continuum photon density ratio undergoes a dramatic reversal in the vicinity of the dust irradiation surface , resulting in a @xmath0-dominated field .
this resurgence appears to more than compensate for the reduction of @xmath0 density due to the h - layer . in the asymptotic limit ( large optical depth ) of the analytic planar model both fields are attenuated according to @xmath100,$]where @xmath101 is the vertical optical depth due to dust , and @xmath102 is a decay constant that depends only on the scattering properties @xmath103 of dust @xcite . as observed , the @xmath0/fuv - continuum ratio approaches a constant value deep inside the disk ( the ` attenuated region ' in figure [ fig : ratios ] ) . at this point both @xmath0 and fuv - continuum fields
are largely isotropized and their transport controlled by dust alone . in effect , the contrasting early life - histories are imprinted into boundary conditions that determine the proportionality coefficients in the asymptotic regime .
a simplified diagrammatical representation of the transport routes is shown in figure [ fig : nh_map ] . as mentioned previously , the dalessio disks to not include separate thermodynamic treatments for gas and dust . in effect
they assume that the two components are thermally coupled .
models show that this coupling breaks down at low densities , causing the gas temperature to rise above that of dust @xcite .
in turn this increases the local scale height of the gas while decreasing its density .
although not explored in this paper , the thermal decoupling will most likely increase the extent of atomic h in the upper layers of the disk .
this is due to two factors ; first , the formation rate of h@xmath7 is density dependent and an increase in the disk scaleheight will be accompanied by a decrease in volumetric density .
second , h@xmath7 formation on grains becomes increasingly inefficient at @xmath104 due to the short time that h atoms spend on grain surfaces before thermally escaping @xcite .
the settling of dust affects the @xmath0 and fuv - continuum fields in two distinct ways .
first , the removal of dust decreases the efficacy of the @xmath5 formation process , resulting in a thicker h - layer ( eqn .
[ eq : h2 ] , fig .
[ fig : columns ] ) .
this ensures that the h - layer processes every stellar @xmath0 incident upon it .
second , the irradiation surface for fuv - continuum photons recedes into the disk as @xmath19 decreases .
the disk becomes flatter and intercepts a smaller fraction of stellar fuv - continuum photons .
a self - consistent disk model that couples disk structure with our new radiative transfer results is required in order to quantify these effects more precisely .
we conclude with a brief summary of the main results : 1 .
the vertical profile of the @xmath0/fuv - continuum photon density ratio is strongly stratified , and may vary by orders of magnitude relative to the intrinsic stellar ratio .
the stratification is strongest in the inner disk .
2 . the radiation field in the vicinity of the disk midplane - where most of the disk mass resides - is relatively enhanced in @xmath0 .
this enhancement is in addition to that already present in the stellar spectrum .
if previous radiative transfer calculations most closely resemble our fuv - continuum results , then the omission of resonantly scattered @xmath0 implies that the mass - averaged intradisk fuv field has been systematically underestimated .
this directly affects estimates of photoelectric heating rates , thermal balance , and the resulting hydrostatic structure .
the settling of dust ( @xmath105 ) lowers the dust irradiation surface and thickens the resonant scattering h - layer by inhibiting @xmath5 formation .
both effects contribute to a more stratified @xmath0/fuv - continuum ratio , ultimately resulting in a greater @xmath0 enhancement in the molecular disk .
the presence of an optically thick , high - albedo h - layer implies that the direct observation of @xmath0 scattered by the disk ( @xmath106 of incident photons ) may be possible in sufficiently close , low extinction systems ( e.g. tw hya ) .
these observations may place constraints on the shape and thickness of the h - layer .
a @xmath0-dominated radiation field drives differential photo - chemical processes in both the gas and grain surfaces @xcite .
numerous pairs of closely - related molecules exhibit a similar differential response to the spectral form of the uv field @xcite . in view of the strongly stratified @xmath0/fuv - continuum ratio one might also expect the photochemical environment to be similarly stratified .
we thank p. dalessio and n. calvet for providing the disk models .
we also thank g. herczeg for providing the @xmath0 spectrum for tw hya .
tb expresses gratitude to members of the idl - pvwave mailing list who provided useful suggestions regarding the computational management of unstructured grids .
the authors acknowledge support from nasa grant nhn08 ah23 g .
markov chain monte carlo simulation is an intuitive and versatile technique for numerically solving high - dimensional , geometrically complex transport problems .
although the method permits many degrees of abstraction , the most intuitive implementation simply involves following the stochastic life - histories of many discrete particles ( in our case , photons or photon packets ) as they travel through a medium , experiencing interactions that are modeled probabilistically @xcite .
if we consider the basic problem of photons being absorbed and scattered then it is primarily the inclusion of scattering that complicates the calculation , and thus motivates the use of the monte carlo method .
additionally , the monte carlo method places essentially no constraints on the geometry involved , provided a meaningful spatial discretization can be obtained .
our monte carlo simulation method is conceptually similar to @xcite .
the reader is referred to this paper for a description of the fundamental steps , which we briefly enumerate here . 1
. the @xmath107 stellar photon packet ( from a total of @xmath108 ) starts at @xmath109 with a luminosity @xmath110 and direction @xmath111 , where @xmath112 is its elevation angle above the midplane .
this is recast into a cartesian direction vector @xmath113 where @xmath114 , @xmath115 , and @xmath116 .
the star is an isotropic light source with total luminosity of @xmath117 .
2 . randomly sampling beer s law generates the optical depth to the next scattering event , @xmath118 , where @xmath119 is a uniformly distributed random number in the range @xmath120 $ ] .
the photon trajectory is propagated through the medium until it has traversed an optical distance @xmath121 , where the opacity @xmath122 for fuv - continuum photons and @xmath123 for @xmath0 photons .
the photon is transported from cell - face to cell - face ( see appendix [ sec : intercept ] ) , each time traversing some distance @xmath124 and optical distance @xmath125 , until the scattering location is overstepped , at which point the photon is regressed back to the exact scattering location .
the optical properties within a cell are considered to be homogeneous .
although the photon is traveling in 3d space , at each face - intercept its coordinates and direction vector are rotated back into the @xmath38 plane .
relevant information ( e.g. packet luminosity @xmath126 ) is deposited as the photon passes through each cell .
there is a continual attrition of photons due to absorption , causing a decrease in the weight - factor @xmath127 w_i$ ] , where @xmath46 is the albedo .
the scattered photon direction is found by performing these steps ; transforming into a frame aligned with the photon direction vector , finding the scattering angle by randomly sampling the scattering phase function , and finally transforming back into the coordinate frame of the disk .
this coordinate transformation is an application of euler angles . in the case of @xmath0 , the type of scattering
must first be resolved .
if a uniformly distributed random variate @xmath128 $ ] satisfies @xmath129 , then we execute resonant scattering , otherwise the scattering is by dust .
repeat from step 2 until photon leaves domain .
the henyey - greenstein phase function used for dust scattering has a number of convenient features that make it attractive for use in monte carlo simulations .
not only can its probability distribution ( eqn [ eq : hg ] ) be integrated over angle to produce a closed - form cumulative probability distribution , but the result can be algebraically inverted , permitting the expression of the scattering angle , @xmath130 , in terms of a uniformly distributed random number , @xmath128 $ ] , @xmath131.\label{eq : hginv}\ ] ] we approximate @xmath0 scattering as isotropic .
this simplifies matters considerably , since the new scattered direction @xmath132 is simply the result of randomly sampling directions over the unit sphere , thus avoiding the computationally expensive euler angle calculation .
we restrict our computational domain to @xmath133 i.e. the volume above the disk midplane .
this is equivalent to asserting a mirror plane at @xmath134 ; if any photons reach the disk midplane they are reflected @xmath135 .
the solution is now smooth and symmetric about the disk midplane , properly accounting for photons that pass vertically from one side of the disk to the other .
when projected onto the @xmath38 plane , straight lines typically appear as hyperbolae .
photons may escape by reaching the outer and upper edges of the domain .
in fact , the delaunay tessellation produces a convex domain perimeter , which in general is not restricted to a rectangular form ..
although the unstructured grid is specified in 2d , we still need to envision the cells as 3d entities in order to use them in the monte carlo radiative transfer simulation .
this is because a photon travels in 3d ( it is not constrained to a single @xmath38 plane ) . by employing an axisymmetric cylindrical framework we are asserting invariance upon rotation about the @xmath136 axis .
once rotated , the triangular simplex is extruded into a torus with a triangular cross - section .
photon trajectories will be propagated through these volumes , requiring geometric tools that allow us to keep track of the intercepts between the photon path and the surfaces of these contiguous volumes .
figure [ fig : cell ] shows how the triangular torus is formed from the interstice of three intersecting cones .
computationally , the most efficient way to transport photons through these volumes is to jump from one cell face ( the entry face ) to another ( the exit face ) . in order to determine these locations we need to calculate the intersections between a straight line ( photon trajectory ) and the three cones that form the surfaces of our toroidal cells ( in 3d ) .
the exit face is always the face with the intersection point that is closest to the current position of the photon ( in the direction of photon propagation ) . which simply states that the vector between a location in space @xmath138 and the cone vertex @xmath139 , makes an angle @xmath140 with the cone axis ( in our case the z - axis , with unit direction vector @xmath141 ) .
we refer to @xmath140 as the opening angle .
it is convenient to square eqn .
[ eq : cone ] to obtain a quadratic version of the cone equation .
there will now be two solutions , one of which corresponds to the reflection cone .
for example , if @xmath138 lies on the cone , then so too does the point @xmath142 that simply corresponds to the reflection of @xmath138 through the vertex @xmath139 .
this formulation automatically deals with cones that are ` upside down ' ( i.e. have opening angle @xmath143 ) . thus , the formulation is comprehensive and it only remains to select the solution corresponding to the appropriate cone orientation .
we proceed by writing the squared eqn [ eq : cone ] in matrix form , where @xmath148 is the unit direction vector , @xmath149 is a point of origin ( the current photon location ) , and @xmath150 the distance along the trajectory .
the intersection between the photon trajectory and cell wall is found by inserting eqn [ eq : line ] into eqn [ eq : matrix ] and solving for the distance @xmath150 . upon doing this
we obtain the quadratic equation , where @xmath152 , @xmath153 , and @xmath154 .
the most case is that @xmath155 , making eqn [ eq : quad ] a quadratic equation with solutions @xmath156 . if @xmath157 no real solutions are possible and no intersection exists
. if @xmath158 a single ( repeated root ) intercept is found , which is seen to be tangent to the cone at the point @xmath159 . if @xmath160 two intercepts exist at @xmath161 .
there are additional possible situations , for example , @xmath162 renders eqn [ eq : quad ] linear , implying that only one intersection exists . in this rather improbable case
the photon trajectory is parallel ( but not coincident ) with a straight line on the cone . finally , if @xmath163 the photon path lies on the cone surface , which is exceedingly unlikely .
a computational implementation should takes these cases into account .
there are extremal cones that may arise from the tessellation and need to be taken into consideration .
if the cell edge in the @xmath38 plane is perpendicular to the z - axis , @xmath164 , then upon rotation we have a horizontal plane rather than a cone .
this can be seen by putting @xmath165 into eqn [ eq : cone ] .
finding the intersection of a line and a plane is straightforward . on the other hand , if the cell edge in the @xmath38 plane is parallel to the z - axis , for example @xmath166 , upon rotation we have a cylinder instead of a cone .
spectrum of tw hya expressed in doppler units , @xmath167 , relative to line center @xcite .
_ bottom _ - the resonant scattering voigt profile ( _ solid line _ ) presented by a gas of h atoms with kinetic temperature @xmath168k .
the lorentzian wings dominate the profile at displacements of @xmath169 .
dust opacity ( _ dashed line _ ) is essentially constant over this small wavelength range , scaling in proportion to the dust abundance @xmath170 [ fig : voigt ] ] , in the @xmath66 modified dalessio disk ( _ filled contours _ ) . a hypothetical atomic hydrogen layer ( ` h layer ' )
is shown by the hatched region .
stellar @xmath0 photons strike the upper surface of this layer .
a fraction of this flux is transmitted , emerging isotropically ( shown by arrows ) .
in contrast , stellar fuv - continuum photons are shown to strike the irradiation surface determined by dust opacity . [
fig : nh_map ] ] wavelength ( @xmath171 , denoted by vertical dotted line ) .
_ middle _ - scattering cross - section .
_ bottom _ - single - scattering albedo , @xmath46 , and asymmetry parameter , @xmath47 . at @xmath34dust has an asymmetry parameter of @xmath35 and albedo @xmath45 .
[ fig : dust_properties ] ] and @xmath37 ( see eqn [ eq : hg ] ) .
isotropic scattering has @xmath42 and is a convenient approximation for modeling the resonant scattering of @xmath0 by h atoms .
the dust ensemble used in this paper has an asymmetry parameter of @xmath35 and is therefore considered to be significantly forward - throwing .
[ fig : hg ] ] calculation in the @xmath66 modified dalessio disk . _ left _ - starting grid comprised of nested regular background arrays and the original dalessio 1 + 1d data .
total number of nodes is @xmath172 . _
middle _ - after one iteration new nodes have been inserted into regions where a change in the h@xmath7 molecular fraction has been detected . _
middle _ - the node distribution after two iterations .
the region requiring refinement becomes smaller as the h / h@xmath7 transition becomes better defined .
the h / h@xmath7 distribution typically converges to a solution after about 15 iterations , at which point the total number of nodes is @xmath173 . in the inner disk
the spatial resolution of the discretization is on the order of [ fig : grid_adaptive ] ] au ) of the discretized @xmath66 modified dalessio disk prior to the h / h@xmath7 calculation . shown are the actual connections between nodes calculated by performing a delaunay tessellation on the node list .
the resulting triangles are the cells through which photon packets propagate . as in figure
[ fig : grid_adaptive ] , the regular array of nodes is the background grid , employing a divide - by-2 scheme to increase the resolution nearer to the star . note that although these nodes are distributed in a rectangular array , the resulting cells generated by the delaunay tessellation are triangular .
the original data from the 1 + 1d dalessio disk model are identified as the vertical columns of nodes disrupting the regularity of the 2d background lattice . in a sense
, the delaunay tessellation heals these discordances , merging disparate sets of nodes .
[ fig : grid_inner ] ] au in the modified dalessio disks with varying degrees of dust settling , @xmath19 .
vertical column densities of atomic hydrogen ( @xmath178 , _ dashed line _ ) and molecular hydrogen ( @xmath64 , _ solid line _ ) , and ( scaled ) volumetric atomic hydrogen density ( @xmath59 , _ dot - dashed line _ ) are shown against the total hydrogen column , @xmath179 .
vertical column densities are integrated downwards from infinity.[fig : columns ] ] ) of the photon density and anisotropy of the @xmath0 ( _ left - hand panels _ ) and fuv - continuum ( _ right - hand panels _ ) radiation fields in the @xmath66 modified dalessio disk model .
the star is located at the origin .
_ top _ - photon density ( units @xmath180 ) .
black arrows indicate the direction of the net flux . _
bottom _ - the anisotropy of the radiation field , @xmath181 , where @xmath132 is the unit direction vector , @xmath182 is the intensity , and @xmath183 is the mean intensity .
the limit @xmath184 implies unidirectionality and @xmath185 implies isotropy .
the flared feature of low anisotropy seen in the @xmath0 panel ( _ lower left _ ) is attributable to ( and coincident with ) the resonantly scattering h - layer .
note also how the net @xmath0 flux emerging from the base of the h - layer does so perpendicularly .
[ fig : maps],scaledwidth=80.0% ] and fuv - continuum photon densities as a function of vertical column density in the modified dalessio disks . _
top _ - at @xmath95au the ratio departs from the intrinsic stellar ratio ( @xmath16 ) as we descend into the disk ( increasing @xmath179 ) .
the asterisks denote the lower surface of the h - layer ( c.f .
[ fig : columns ] ) .
_ bottom _ - the same at @xmath186au .
the shaded areas provide an estimate of the monte carlo noise .
the curves are terminated at the depth where the noise begins to degrade the data significantly .
interpretative text has been added for the @xmath186au , @xmath66 case . [
fig : ratios],scaledwidth=65.0% ] and fuv - continuum photon densities as a function of vertical column density in the modified dalessio disks .
_ top _ - at @xmath95au the ratio departs from the intrinsic stellar ratio ( @xmath16 ) as we descend into the disk ( increasing @xmath179 ) .
the asterisks denote the lower surface of the h - layer ( c.f .
[ fig : columns ] ) .
_ bottom _ - the same at @xmath186au .
the shaded areas provide an estimate of the monte carlo noise .
the curves are terminated at the depth where the noise begins to degrade the data significantly .
interpretative text has been added for the @xmath186au , @xmath66 case . [
fig : ratios],scaledwidth=65.0% ] .
the three extended lines that form the cell edges intercept the z - axis at points @xmath187,@xmath188 and @xmath189 . upon rotation about the z - axis
these lines become the surfaces of three cones with vertices at @xmath187 , @xmath188 and @xmath189 .
the cone with vertex @xmath187 is shown by the dashed lines .
the interstice formed by the three intersecting cones is a toroid with triangular cross - section .
[ fig : cell],scaledwidth=65.0% ] | we study the role resonant scattering plays in the transport of @xmath0 photons in accreting protoplanetary disk systems subject to varying degrees of dust settling . while the intrinsic stellar fuv spectrum of accreting t tauri systems may already be dominated by a strong , broad @xmath0 line ( @xmath180% of the fuv luminosity )
, we find that resonant scattering further enhances the @xmath0 density in the deep molecular layers of the disk .
@xmath0 is scattered downwards efficiently by the photodissociated atomic hydrogen layer that exists above the molecular disk .
in contrast , fuv - continuum photons pass unimpeded through the photodissociation layer , and ( forward-)scatter inefficiently off dust grains . using detailed , adaptive grid monte carlo radiative transfer simulations
we show that the resulting @xmath0/fuv - continuum photon density ratio is strongly stratified ; fuv - continuum dominated in the photodissociation layer , and @xmath0-dominated field in the molecular disk .
the enhancement is greatest in the interior of the disk ( @xmath2au ) but is also observed in the outer disk ( @xmath3au ) .
the majority of the total disk mass is shown to be increasingly @xmath0 dominated as dust settles towards the midplane . |
solid - state lasers containing frequency - doubling crystals are efficient and compact sources of coherent visible optical radiation . unfortunately , when they operate in multimode regime , one observes irregular fluctuations of the output intensity .
this behavior , referred to as the green problem , has been reported for the first time by baer [ baer , 1986 ] .
he found that these instabilities arise from a coupling between longitudinal modes of the laser due to sum - frequency generation .
in particular , when such a laser operates in a single longitudinal mode , its output is stable [ kennedy & barry , 1974 ] . in the case of two oscillating longitudinal modes , the output intensity is stable only for small values of nonlinearity , otherwise both modes tend to pulse on and off out of phase [ baer , 1986 ] .
when the number of lasing modes is larger than two , the laser can exhibit , depending on the parameters describing it , various types of behaviour such as : antiphase dynamics [ james et al .
, 1990b , wiesenfeld et al.,1990 , james et al .
, 1990a , roy et al . , bracikowski & roy , 1991 , mandel & wang , 1994 , wang et al . , 1995 ,
otsuka et al . ,
1997 ] , clustering [ baer , 1986 ] , grouping [ otsuka et al . , 1997 ] , and chaotic dynamics [ james et al .
, 1990a , roy et al .
, bracikowski & roy , 1991 , liu et al . ,
1997 ] .
one of the motivations of the large number of papers devoted to intracavity - doubled solid state lasers and published during the last decade has been the need to find a configuration and the range of parameters describing the lasers , which could guarantee a stability of their outputs .
several ways to reach this goal have been found , for example , an introduction of a quarter - wave plate into the laser cavity , a proper alignment of the angle between birefringent axes of the active medium and the nonlinear crystal [ james et al .
, 1990a , james et al .
, 1990b , roy et al . ] , a placing of a tilted mirror inside the cavity [ ustyugov et al . , 1997 ] , or an utilization of the effect of sum - frequency - mixing [ falter et al . , 1997 ,
danailov & apai 1994 ] .
moreover , it has been predicted analytically [ james et al . , 1990b ] and confirmed experimentally [ magni et al . , 1993 ]
that when the number of oscillating longitudinal modes is set to be sufficiently large ( for example , by increasing the length of the laser cavity ) the output of the laser becomes stable . besides
, the green problem can also be avoided in the case of a single mode operation of the laser , which can be obtained , for example , by use of an intracavity etalon [ baer , 1986 ] or a birefringent filter [ nagai et al . , 1992 , fan , 1991 ] the main goal of this paper is to continue the investigations of the green problem .
the fact that the cross - saturation coefficient is modulated by the spatial hole - burning effect is taken into account .
firstly , the stabilization of the laser radiation by an increase of the number of longitudinal modes , as proposed in [ james et al . , 1990b , magni et al . , 1993 ] is analyzed .
the results presented in [ james et al . , 1990b , magni et al . ,
1993 ] , where the cross - saturation coefficient was assummed to be constant for all modes , are compared with the numerical data .
it is shown that the theoretically obtained [ wang & mandel , 1993 ] linear dependence of the minimal number of modes necassary for stabilization the laser output on the strength of nonlinearity agrees with the numerical solutions only in the case of sufficiently small nonlinearity . for larger values of nonlinearity , due to the cancellation of modes during the evolution , the minimal number of modes obtained by the numerical simulations is larger than the number which follows from the theoretical predictions . for very large nonlinearity
this cancellation is so strong that only a few modes survive ( even when there are initially 250 oscillating modes in the laser cavity ) .
therefore , a large number of simultaneously oscillating modes necessary for stabilization of the laser output can not be achieved .
a similar situation takes place when the cross - saturation coefficient is modulated by the effect of spatial hole - burning , therefore the stabilization of the laser output by an increase of the number of longitudinal modes is not always possible , as a result of a strong competition between modes and a cancellation of some of them during the evolution .
however , the problem of the stabilization of the laser output can be solved in another way , namely , by an increase of the strength of nonlinearity , which leads to very strong competition between the modes , so that during the evolution all of them , but a single one , are canceled .
as a consequence , a steady - state solution , which is stable against small perturbations , arises .
this is the stabilization mechanism proposed in the present paper .
it is valid in the case of constant cross - saturation coefficient , as well as in the case when the spatial hole - burning is taken into account .
the analysis presented in this paper is based on the baer - type rate equations [ baer , 1986 ] extended by roy , bracikowski and james to the case when the effect of spatial hole - burning is taken into account [ roy et al . ,
bracikowski & roy , 1991 ] .
these equations have the following form : @xmath0 ( p , q ) = _ 0 _ 0 | |_z_0^z_0+l , [ cross ] @xmath1 where @xmath2 is the number of longitudinal modes ; @xmath3 and @xmath4 are , respectivelly , the cavity round trip time and the fluorescence lifetime ; @xmath5 and @xmath6 are , respectively , the intensity and gain of the @xmath7-th longitudinal mode ; @xmath8 is the cavity loss parameter for the @xmath7-th mode ; @xmath9 is the small signal gain ; @xmath10 and @xmath11 describe , respectivelly , self - saturation of the @xmath7-mode and cross - saturation between two modes , @xmath7 and @xmath12 ; @xmath13 is the scaling parameter ; @xmath14 is the length of the gain medium ; @xmath15 is the length of the laser cavity , @xmath16 is the distance of the gain medium from the first cavity mirror , ( as shown in fig .
1 ) ; @xmath17 is the wavevector of the longitudinal cavity mode @xmath7 with the wavelengths @xmath18 .
note that even through in eq .
( [ cross ] ) three parameters : @xmath15 , @xmath14 , and @xmath16 are present , the modified cross - saturation coefficient depends only on two rescaling parameters , @xmath19 and @xmath20 .
the parameter @xmath21 is a nonlinear coefficient which describes the conversion efficiency of the fundamental intensity into the doubled intensity
. the terms @xmath22 and @xmath23 in eq .
( 1 a ) account for the loss in the intensity of the fundamental frequency through second harmonic generation and through sum - frequency generation , respectivelly .
the set of equations ( 1 a , b ) has a rather complicated structure and it is not possible to solve it analytically .
therefore , the results presented in this paper has been obtained with the aid of the numerical method of runge - kutta .
in this section the approximation that the cross - saturation coefficient is constant for all modes : @xmath24 , is used .
it is assumed that losses and small signal gains are the same for all modes , i.e. @xmath25 , where @xmath26 .
other parameters describing the system have been chosen as follows : @xmath27 , \t_f=0.24 \ , [ ms ] , \al=0.015 , \g=0.12 , \b_0=0.06 $ ] .
the number of longitudinal modes , @xmath28 , and the strength of nonlinearity , @xmath29 , are not fixed and they are varied in the analysis .
as the first step let us consider the dependence of the laser output on the number of longitudinal modes initially excided in the laser cavity . from the numerical simulations it follows that for very small nonlinearity , @xmath30 , the laser output is stable , even for a large number of longitudinal modes . in the case of larger nonlinearity , @xmath31 , as it can be seen from fig .
2a , the behavior of the laser output is complicated and already for three simultaneously oscillating longitudinal modes the total intensity exhibits irregular oscillations .
when the number of modes increases , amplitudes of those oscillations slowly increase ( see fig .
2b representing the results for 70 modes ) .
finally , when the number of modes is larger than the critical value , the total stabilization of the laser output occurs , as shown in fig .
however , for slightly higher than previously considered value of nonlinearity , @xmath32 , we have not observed the stabilization of the laser output , even for very large number of longitudinal modes , @xmath33 ( compare the total intensity of the laser output for @xmath34 , @xmath35 , and @xmath33 modes , shown , respectivelly , on figs . 2d , 2e , and 2f ) .
this happens because of cancelation of modes , as a result of a competition between them , which takes place during the evolution . the dependence of the minimal number of modes neceessary for stabilization of the laser output on the strength of nonlinearity is presented in fig .
3 . as it can be seen , for small nonlinearity , @xmath36
, there is a rather good agreement between theoretically [ wang & mandel , 1993 ] and numerically obtained results .
however , when nonlinearity increases , @xmath37 , the minimal number of modes obtained in numerial simulations is larger than the number predicted theoretically .
this discrepancy is the result of the competition between modes , which leads to cancellation of some of them during the evolution .
when @xmath38 the cancellation is so big that after some time only few modes survive , thus a sufficiently large number of modes necessary for stabilization of the laser can not be realised .
concluding , depending on the relation between nonlinearity and the number of longitudinal modes , the dynamics of a multimode , intracavity - doubled , solid - state laser can be divided into four regions as illustrated in fig .
5 . in particular , when nonlinearity is small and the number of modes is larger than a critical value the output of the laser is stable .
examples of such a behavior , which is labeled as region i , are shown in fig .
4a . with increasing nonlinearity the complexity of the behaviour of the laser increases and irregular fluctuations of the output intensity appear .
this behavior , depicted in fig .
4b , corresponds to region ii . for even larger values of the strength of nonlinarity the cancellation of modes
starts to take place .
when nonlinearity exceedes the critical value @xmath39 the evolution of the output intensity leads to a cancellation of a large number of modes , so that only a few of them survive .
as s result , a stabilization of the laser output can not be obtained .
this behavior , illustrated in fig .
4c , is classified as region iii . when nonlinearity is larger than the critical value @xmath40 there is such a large mutual cancellation of modes that after some time only one of them survives .
therefore , the steady - state solution , shown in fig .
4d , which is stable against small perturbations , arises .
this mechanism of stabilization , achieved by forcing the laser to operate in the one - mode regime , is similar to other approaches presented in the literature , where the stabilization is obtained by inserting into the laser cavity an additional element like an etalon [ baer 1986 ] or a birefringent crystal [ nagai et al . , 1992 , fan , 1991 ] .
however , the method proposed here seems to be a better solution , since no additional element is involved .
in this section the analysis presented in the previous section is extended to take into account the effect of spatial hole - burning .
the present analysis is based on the baer - type rate equations , eq .
( 1 ) , modified by bracikowski , roy and james [ roy et al . ,
bracikowski & roy , 1991 ] , who have written down an explicit form for the cross - saturation coefficient , eq .
( 2 ) , but were not investigating the influence of the spatial hole - burning on the laser dynamics .
as far as the autors know , the analysis presented in this section is the first study of this issue .
similarly to the previous section , two ways of the laser stabilization are analyzed : by increasing the number of longitudinal modes and by increasing the strength of nonlinearity .
parameters describing the laser cavity are chosen to be as follows : @xmath41,$ ] @xmath42,$ ] @xmath43 $ ] .
our results show that when the spatial hole - burning effect is taken into account ( i ) the competition between the modes is much stronger than in the case of the cross - saturation coefficient constant for all modes ( ii ) .
therefore , the minimal number of modes which are necessary to obtain the stabilization of the laser output is in the case ( i ) larger than in the case ( ii ) .
for example , for the strength of nonlinearity @xmath44 only 80 modes are necessary in the case ( ii ) , as shown on fig .
2c , while in the case ( i ) even 200 initially oscillating modes are not sufficient for the stabilization of the laser output , as shown on fig . 6c . therefore , an alternative approach of the stabilization of the laser output based on an increasing of the strength of nonlinearity , which has been proposed in the previous section , can be examined . indeed , from the results of the numerical simulations , displayed in fig .
6f it follows that the stabilization of the laser output can be obtained when the nonlinearity is larger than a critical value .
this occurs because of an extensive cancellation of modes , due to which the laser operates in a single longitudinal mode .
in this paper the properties of the dynamics of multimode intracavity - doubled solid - state lasers have been studied .
the infulence of spatial hole - burning on the cross - saturation coefficient was taken into account .
the system was described by the rate equations of baer - type , which were solved numerically with the aid of the runge - kutta method .
firstly , the case of the cross - saturation coefficient constant for all modes was considered .
the minimal number of modes necassary for the stabilization of the laser output was defined and its dependence on the strength of nonlinearity was studied .
it was shown that for small nonlinearity this dependence is linear and agrees with the theoretical predictions . however , for larger values of nonlinearity the minimal number of modes was found to be larger than the number obtained in theoretical considerations . as a reason of this discrepancy the effect of the competition between the modes and cancellation of some of them during the evolution was given .
it was also observed that with increasing of nonlinearity the competition between modes increases , so that for some values of nonlinearity the initially large number of oscillating modes is reduced to only a few of them . for much higher values of nonlinearity
this competition is so strong that only a few modes survive .
therefore , a large number of simultaneously oscillating modes can not be realised and the stabilization can not be obtained in this way .
however , we propose another method based on the increase of the strength of nonlinearity .
this stabilization occurs for such a high value of nonlinearity , for which all modes , besides a single one , are canceled during the evolution . in this case
a steady - state solution , stable against small perturbations , arises .
a similar analysis was accomplished for the cross - saturation coefficient modulated by the effect of spatial hole - burning .
we have shown that in this case the stabilization of the laser output by increase of the number of longitudinal modes is hardly possible .
therefore , an alternative approach of the stabilization based on an increasing of the strength of nonlinearity have been examined and shown to be an appropriate solution .
the method of a stabilization of a laser output , as proposed in our paper , is similar to other approaches presented in the literature where the stable output of the laser is achieved by forcing the laser to operate in a single - mode regime , for example , by inserting into the cavity an etalon or a birefringent crystal .
however , the method proposed here offers a better solution , since no insertion of an additional element into the laser cavity is needed .
the authors are grateful to prof .
m. virasoro , the director of the abdus salam international centre for theoretical physics , trieste ( italy ) , for granting this work and to prof .
g. denardo for his encouragement ; m. d. wishes to acknowledge the support of the international centre for science and high technologies ( i.c.s . ) , trieste ( italy ) .
* baer , t. [ 1986 ] `` large - amplitude fluctuations due to longitudinal mode coupling in diode - pumped intraavity - doubled nd : yag lasers , '' _ j. opt .
b _ * 3 * , 1175 . * c. bracikowski , c. & roy , r. [ 1991 ] `` chaos in a multimode solid - state laser system , '' _ chaos _ * 1 * , 49 . * danailov , m. b. & apai , p. [ 1994
] `` 589 nm generation by intracavity mixing in a nd : yag laser , '' _ j. appl .
phys . _ * 75 * , 8240 . *
falter s. et al . [ 1997 ] `` dynamics and stability of a laser system with second - order nonlinearity , '' _ opt .
* 22 * , 609 . *
fan , t. y. [ 1991 ] `` single - axial mode , intracavity doubled nd : yag laser , '' _ ieee j. q. e. _ * eq-27
* , 2091 . *
james , g. e. et al . [ 1990 ] `` intermittency and chaos in intracavity doubled lasers .
ii , '' _ phys .
rev . a _ * 41 * , 2778 .
* james , g. e. et al . [ 1990 ] `` elimination of chaos in an intracavity - doubled nd : yag laser , '' _ opt .
lett . _ * 15 * , 1141 . * kennedy , c. j. & barry , j. d. [ 1974 ] `` stability of an intracavity frequancy - doubled nd : yag laser , '' _ ieee j. q.e _ * qe-10 * , 596 . * liu , c. et al .
[ 1997 ] `` influence of noise on chaotic laser dynamics , '' _ phys .
e _ * 55 * , 6483 . * magni , v. et al .
[ 1993 ] `` intracavity frequency doubling of a cw high - power tem00 nd : ylf laser , '' _ opt .
* 18 * , 2111 . *
mandel , p. & wang , j. [ 1994 ] `` dynamic independence of pulsed antiphased modes , '' _ opt .
* 19 * , 533 . * nagai , h. et al .
[ 1992 ] `` low - noise operation of a diode - pumped intracavity - doubled nd : yag laser using a brewster plate , '' _ ieee j. q. e. _ * qe-28 * , 1165 . *
otsuka , k. et al .
[ 1997 ] `` clustering , grouping , self - induced switching , and controlled dynamic pattern generation in a antiphase intracavity second - harmonic generation laser , '' _ phys .
e _ * 56 * , 4765 .
* roy , r. et al . [ 1993 ] `` dynamics of a multimode laser with nonlinear , birefringent intracavity elements , '' in _ recent developments in quantum optics _ , ed .
inguva , r. ( plenum press , new york ) .
* ustyugov , v. i. et al .
[ 1997 ] `` self - stabilization od a continuous wave multimode green laser due to antiphase dynamics , '' _ app .
_ * 71 * , 154 .
* wang , j - y . & mandel , p. [ 1993
] `` antiphase dyamics of multimode intracavity second - harmonic generation , '' _ phys .
a _ * 48 * , 671 . * wang , j - y . at al . [
1995 ] `` antiphase states in intracavity second harmonic generation , '' _ quant .
semiclass . opt . _ * 7 * , 169 . *
wiesenfeld , k. et al .
[ 1990 ] `` observation of antiphase states in a multimode laser , '' _ phys .
* 65 * , 1749 .
[ 2cm][0 cm ] @xmath45 @xmath45 + [ 1cm][-1 cm ] a)@xmath46 \hspace{4 cm } d ) $ ] @xmath47 $ ] + [ 2.0cm][4.0 cm ] + [ 0cm][0 cm ] @xmath45 @xmath45 + [ 1cm][-1 cm ] b)@xmath46 $ ] e ) @xmath47 $ ] + [ 2.0cm][4.0 cm ] [ 0cm][0 cm ] @xmath45 @xmath45 + [ 1cm][-1 cm ] c)@xmath46 $ ] f ) @xmath47 $ ] + [ 2.0cm][1.0 cm ] [ fig2 ] [ 2cm][0 cm ] @xmath45 @xmath45 + [ 1cm][-1 cm ] a)@xmath46 \hspace{4 cm } d ) $ ] @xmath47 $ ] + [ 2.0cm][4.0 cm ] + [ 0cm][0 cm ] @xmath45 @xmath45 + [ 1cm][-1 cm ] b)@xmath46 \hspace{4 cm } e ) $ ] @xmath47 $ ] + [ 2.0cm][4.0 cm ] [ 0cm][0 cm ] @xmath45 @xmath45 + [ 1cm][-1 cm ] c)@xmath46 \hspace{4 cm } f ) $ ] @xmath47 $ ] + [ 2.0cm][1.0 cm ] [ fig11 ] | in this paper possibilities of a stabilization of large amplitude fluctuations in an intracavity - doubled solid - state laser are studied .
the modification of the cross - saturation coefficient by the effect of spatial hole - burning is taken into account .
the stabilization of the laser radiation by an increase of the number of modes , as proposed in [ james et al .
, 1990b , magni et al . , 1993 ] , is analyzed .
it is found that when the cross - saturation coefficient is modulated by the spatial hole - burning the stabilization is not always possible .
we propose a new way of obtaining a stable steady - state configuration based on an increase of the strength of nonlinearity , which leads to a strong cancellation of modes , so that during the evolution all of the modes , but a single one , are canceled .
such a steady - state solution is found to be stable with respect to small perturbations .
16.5 cm 23.cm -1.5 cm
i i v elimination of chaos in multimode , intracavity - doubled lasers in the presence of spatial hole - burning .
+ faculty of physics , warsaw university of technology + warsaw , poland + + laser laboratory , sincrotrone + trieste , italy + short title : elimination of chaos in multimode , intracavity - doubled ... |
it is well understood that fractional - order derivatives provide a good tool for the description of memory and hereditary properties of various processes , fractional - order systems being characterized by infinite memory .
generalizations of dynamical systems using fractional - order derivatives instead of classical integer - order derivatives have proved to be useful and more accurate in the mathematical modeling of real world phenomena arising from several interdisciplinary areas such as : diffusion and wave propagation , viscoelastic liquids , fractional kinetics , boundary layer effects in ducts , electromagnetic waves , electrode - electrolyte polarization
. theoretical characterization of chaos in fractional - order dynamical systems is yet to be investigated .
however , chaotic behavior has been observed by numerical simulations in many systems such as : a fractional - order van der pol system @xcite , fractional - order chua and chen s systems @xcite , a fractional - order rossler system @xcite and a fractional - order financial system @xcite .
nevertheless , it is worth noting that numerical simulations are limited by the fact that they only reveal the chaotic behavior of discrete - time dynamical systems that are obtained by discretizing the fractional - order systems . in order to assess chaotic behavior of fractional - order dynamical systems ,
it is extremely important to be able to accurately estimate the solutions over a large time interval .
several numerical methods are used for fractional - order systems , such as a generalization of the adams - bashforth - moulton @xcite predictor - corrector method or a class @xcite of p - fractional linear multistep methods . for the numerical assessment of chaotic behavior of fractional - order systems
, it is extremely important to accurately estimate the solutions over a large time interval . in numerical schemes such as the adams - bashforth - moulton predictor - corrector method , due to the hereditary nature of the problem and the non - locality of fractional - order derivatives , at every iteration step all previous iterations have to be taken into account
. it may be feasible to overcome these difficulties with the aid of parallel computing algorithms ( e.g. @xcite ) implemented in a conventional way or using available high performance computing systems . in this paper we present the case of adams - bashforth - moulton predictor - corrector method and measure the speedup of two parallel approaches by using gpu and hpc implementations . in the following
, we will present the theoretical problem , we will describe the parallel dynamic scheme and then we will present our simulation results and our conclusions .
were trying to solve an ordinary fractional differential equation of the form : @xmath0\\ y^{(k)}(0 ) = y_0^{k } , & k \in \{0 , \dots , { \left\lceil \alpha \right\rceil } - 1\ } , \end{cases}\ ] ] where @xmath1 $ ] and @xmath2 denotes the ceiling function that rounds up to the nearest integer .
the fractional derivative is defined as : @xmath3 the numerical method used to solve is a fractional version of the adams - bashforth - moulton method ( a classic predictor - collector method ) presented also in figure [ alg : abm ] .
the domain @xmath4 $ ] is discretized into @xmath5 intervals with a step size @xmath6 and the grid points @xmath7 , for @xmath8 .
we will also denote @xmath9 and @xmath10 with @xmath11 as the initial condition .
since we restrict @xmath12 to the @xmath13 $ ] interval , we do not require any of the higher derivatives of the initial condition .
the first step of the scheme is the * predictor * , which will give a first approximation @xmath14 of our solution : @xmath15 where the weights @xmath16 are given by : @xmath17 the second and final approximation of our solution , called the * corrector * , is given by : @xmath18 where the weights @xmath19 and @xmath20 are defined as : @xmath21 and @xmath22 the most difficult part of numerically solving such equations consists in the fact that at each step , we require the complete history of the variable . that is to say , to compute @xmath23 , we need to know all the previous values @xmath24 that are used to compute @xmath25 , for @xmath26 .
this makes fractional differential equations notoriously hard to parallelize .
some attempts have been made , see @xcite for a new method created specifically for fractional differential equations or @xcite for an adaptation of the _ parareal _ @xcite method to this kind of equations .
the first algorithm we are going to look at proposes to break up all the @xmath5 time steps between the @xmath27 processes we have , giving : @xmath28 steps per process ( see figure [ fig : partition ] ) ( 0 , 0 )
( 3 , 0 ) ; in 0 , 0.5, ...
,1 ( , -0.05 )
( , 0.05 ) ; at ( , -0.05 ) [ below ] @xmath29 ; in 2 , 2.5, ... ,3 ; ( , -0.05 )
( , 0.05 ) ; at ( , -0.05 ) [ below ] @xmath30 ; ( 0,-0.4 ) ( 1,-0.4 ) ; ( 2,-0.4 ) ( 3,-0.4 ) ; at ( 0.5 , -0.6 ) [ below ] process @xmath31 ; at ( 2.5 , -0.6 ) [ below ] process @xmath32 ; at ( 1.5 , -0.2 ) @xmath33 ; during an iteration @xmath34 , a process @xmath35 that represents the block in which we can find @xmath34 , that is to say if @xmath36\ ] ] will compute the value of @xmath37 .
this is done by computing a partial sum in each process @xmath38 and gathering the results to finally form the complete sum as defined in .
one major downside of the algorithm we are going to describe is that given a fix number @xmath27 of processes , most of them will be idle for a big part of the computations . to be exact , they will be idle as long as : @xmath39 that is to say , as long as the current iteration has not reached the block corresponding to process @xmath35 .
= = = by 2 by -2 bywidthheight 0pt depth 0pt
on our tests we used uvt s bluegene / p cluster that consists of an fully loaded single bluegene / p rack that has move than 1000 cpus and 4 tb of ram memory . it can offer a 11.7 tflops sustained performance .
one important issue of the algorithm described previously ( algorithm [ alg : abm ] ) is that for a fix number @xmath27 of processes , most of them will be idle for the time when the sum is computed from the partial results that are gathered .
below are several simulation to find a balance for this drawback .
we made several simulation of computing a number of elements , with different number of processes , and we observe the average running time for different time computation per process : .bluegene simulation [ cols= " > , > , > , > , > " , ] [ table : avgruntime ] the parallel algorithm for the adams - bashforth - moulton scheme using mpi ( algorithm 1 ) was been also adapted for cuda .
the parabm algorithm in this case suffered few changes : the partial sums are computed using parallel sum reduction ; the weights @xmath19 , @xmath40 , @xmath20 not being dependent on terms being computed in advance via previous steps , they can be preprocessed in parallel , taking into consideration that each thread n computes the weight at time n. in table [ table : avgruntime ] we compare all together the absolute average time of multiple gpu cores versus multiple cpus .
we can observe in figure [ fig : hpcvsgpu ] that in both cases the absolute processing time is increasing with number of steps .
it is surprising that we obtained on hpc cpu a lower rate versus gpu , that tell us we should look carefully to the hpc processing time and idle time on the future .
[ h ] in both approaches we simulate the three - dimensional hindmarsh - rose model @xcite and we observed ( figure [ fig : numerical ] ) that is important to have a big number of steps in numerical simulation because the solution stability can not be seen even after a huge number of iterations , which could take very long time to simulate without parallel processing .
we observed that fixing the number of processes that we need for a run is a huge problem which can be solved by adding new processes as the iteration progress or with iterative algorithms that does not start all the treads / processes at the very early beginning . using cuda to simulate such an algorithm was not a bad idea , however computing at each step of evolution the convolution for the entire history needs complicated reductions algorithms , which should be investigated closely in the future .
we present here few results from our work in progress , as a future work we should continue to investigate in both hpc and gpu approaches , what is the best way to find the numerical solution for such fractional - order systems .
further simulations would possible give us a way for the best compromise in both cases hpc od gpu .
this research has been supported by cncs idei grant pn - ii - ru - te-2014 - 4 - fradys : `` theoretical and numerical analysis of fractional - order dynamical systems and applications '' . | in this paper , we explore how numerical calculations can be accelerated by implementing several numerical methods of fractional - order systems using parallel computing techniques .
we investigate the feasibility of parallel computing algorithms and their efficiency in reducing the computational costs over a large time interval . particularly , we present the case of adams - bashforth - mouhlton predictor - corrector method and measure the speedup of two parallel approaches by using gpu and hpc cluster implementations .
* keywords : * fractional - order systems , parallel numerical algorithms , gpu processing , hpc processing |
a useful question to ask when one studies quantum fields in general relativity is the following : given that all matter is inherently quantum in nature , will quantum effects remove the singularity at the centre of a black hole spacetime ? to answer this question using a scalar field and semi - classical perturbation theory , one must first calculate the expectation value @xmath1 where @xmath2 is the stress - energy tensor operator of the scalar field @xmath3 . in this paper @xmath0 is computed in preparation for the computation of @xmath1 which is used as the source term for the einstein field equations : @xmath4 a review of handling quantum fields in the presence of strong gravitational fields can been found in articles by wipf @xcite and dewitt @xcite and in books by birrel and davies @xcite and wald @xcite .
@xmath0 for massless fields has been computed for both the interior and exterior of a schwarzschild black hole @xcite @xcite .
these calculations have also been extended by anderson to accommodate massive fields in general spherically symmetric , asymptotically flat spacetimes @xcite . a method has also been developed by anderson , hiscock and samuel @xcite @xcite to calculate the expectation value of the stress - energy operator , @xmath1 in spherically symmetric , static spacetimes .
they use this method to calculate @xmath1 in schwarzschild and reissner - nordstrm geometries .
@xmath1 has also previously been computed by howard and candelas @xcite in schwarzschild spacetime .
the kerr spacetime has also been studied with fields propagating in this geometry .
frolov @xcite has calculated @xmath0 for massless fields on the event horizon pole of a kerr - newmann black hole as well as deriving an approximate expression for @xmath1 near the horizon with thorne @xcite .
massive fields in the exterior geometry have been studied by frolov and zelnikov @xcite .
various works on back reaction effects of quantum fields on black hole geometries have been produced .
for schwarzschild and reissner - nordstrm spacetimes this includes the work of hiscock and weems @xcite , bardeen and balbinot @xcite @xcite , and york @xcite who used page s analytic approximation @xcite for @xmath1 in einstein spacetimes for conformally invariant fields .
more recently hiscock and larson @xcite have extended their analysis to the schwarzschild interior and calculated back reaction effects on curvature invariants .
most work in this field has been done in the context of spherical or oblate symmetry .
we wish to extend the analysis to other symmetries and ultimately ask if the above results are general or are specific to the particular symmetry chosen .
for example , does the presence of the quantum field have the same effect on the curvature growth and on the anisotropy of a black hole interior for all symmetries ? if curvature invariants are weakened for all systems studied then one can say with some confidence that quantum effects may remove the singularity .
the system studied here will be that of a massless klein - gordon field with conformal curvature coupling propagating in the spacetime generated by a straight black string .
this background is chosen for several reasons .
first , it possesses cylindrical ( as opposed to spherical ) symmetry and secondly , in the context of cosmic strings it represents a system which may physically exist in the universe .
it could also be argued that sufficiently close to a black string loop the spacetime will possess this type of geometry .
there has also been a revived interest in anti - desitter spacetimes in the context of conformal field theories .
the metric chosen is that developed by kaloper @xcite and lemos and zanchin @xcite which admits a line element of the following form : @xmath5d\varphi^2 \nonumber \\ & + \frac{d\rho^2}{\alpha^2\rho^2- \frac{2(3\omega - m)}{\alpha\rho}+\frac{(3\omega - m)4q^{2 } } { ( \omega+m)\alpha^2\rho^2 } } + \alpha^2\rho^2\ ; dz^2 \label{eq : genmetric}\end{aligned}\ ] ] where @xmath6 , @xmath7 , and @xmath8 are the mass , charge , and angular momentum per unit length of the string respectively .
@xmath9 is given by : @xmath10 the constant @xmath11 is defined as follows : @xmath12 with @xmath13 , the cosmological constant , negative giving the spacetime its asymptotically anti - desitter behaviour .
the spacetime has a well defined time - like killing vector field with respect to which modes can be defined . much interesting
work has been done regarding effects of quantum fields in the 3d btz @xcite black hole @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
the system here however does not dimensionally reduce to the 3d btz black hole whose four dimensional counterpart ( without charge or angular momentum ) is given for reference as @xcite @xmath14 this is due to the fact that the dilaton field is non zero and non constant in the corresponding three dimensional action . in the spacetime considered here ,
both charge and angular momentum are zero yielding the following for ( [ eq : genmetric ] ) @xmath15 it can immediately be seen from ( [ eq : metric ] ) that the spacetime behaves as anti - desitter as @xmath16 and therefore the spacetime is not globally hyperbolic .
this fact requires boundary conditions to be imposed to control the flow of information in and out of the time - like infinity .
it has been shown @xcite that three natural boundary conditions arise in ads@xmath17 spacetime and the condition used here is that of a `` transparent '' boundary condition .
in this section a euclidean space approach is used to calculate the green function .
the calculation is similar to that done by anderson @xcite who calculated @xmath0 in a reissner - nordstrm spacetime however , the method is extended here to a system which is neither spherically symmetric nor asymptotically flat .
the euclidean space method amounts to making the transformation @xmath18 in metric ( [ eq : metric ] ) .
the quantity @xmath0 is then defined as @xmath19 where @xmath20 is the euclidean space green function satisfying the equation @xmath21g_{e}(x , x')= \frac{-\delta^{4}(x , x')}{\sqrt{g(x)}}. \label{eq : kg}\ ] ] the mass term in the klein - gordon operator will be set to zero in order to comply with the transparent boundary conditions mentioned earlier .
the presence of three killing fields allows the right - hand - side of the above equation to be expanded in terms of cylindrical functions as follows : @xmath22 with @xmath23 where @xmath24 is the temperature of the field . the green function
can similarly be expanded giving @xmath25 the function @xmath26 must satisfy the following equation @xmath27=\frac{-\delta(\rho,\rho')}{\rho^2\alpha}. \label{eq : evolve}\end{aligned}\ ] ] a solution is assumed of the form @xmath28 where @xmath29 and @xmath30 represent the lesser and greater of @xmath31 and @xmath32 respectively . integrating across the @xmath33 function gives the wronskian normalisation condition : @xmath34 = \frac{-1}{\alpha\rho^{2}(\alpha^2\rho^2-\frac{4m}{\alpha^2\rho^2})}. \label{eq : wronsk}\ ] ] the asymptotic behaviour of the solution can be found by studying the solutions to ( [ eq : evolve ] ) in the appropriate regimes . for @xmath35
the solution has the form : @xmath36 whereas near the horizon @xmath37 solutions are found to behave as : @xmath38 where @xmath39 .
it can easily be seen that @xmath40 diverges at infinity and @xmath41 is divergent at the horizon . the general solution with the correct asymptotic behaviour
can be found by ansatz .
such a solution takes the form @xmath42 @xmath43 is a function of @xmath31 which evolves according to the equation : @xmath44 \nonumber \\ & + & \frac{9}{4}-\frac{36m^2}{\alpha^6\rho^6 } + \frac{\xi r}{\alpha^2}\left(1-\frac{4m}{\alpha^3\rho^3}\right ) + \frac{x'}{x}\left[\frac{1}{2}\rho+\frac{2m}{\alpha^3\rho^2 } -\frac{16m^2}{\alpha^6\rho^5}\right ] \label{eq : x } \\ \label{eq : xevolve } & + & \left(\frac{x'}{x}\right)^{2}\left[-\frac{3}{4}\rho^2 + \frac{6m}{\alpha^3\rho}-\frac{12m^2}{\alpha^6\rho^4}\right ] + \frac{x''}{x}\left[\frac{1}{2}\rho^2-\frac{4m}{\alpha^3\rho}+ \frac{8m^2}{\alpha^6\rho^4}\right ] .
\nonumber \end{aligned}\ ] ] which can be obtained by substituting ( [ eq : ansatz ] ) into ( [ eq : kg ] ) . substituting ( [ eq : ansatz ] ) into ( [ eq : wronsk ] ) gives @xmath45 for all mode functions .
a point splitting algorithm developed by christensen @xcite will be used to renormalise the field . in this technique
, one chooses the points @xmath46 and @xmath47 to be nearby points in the spacetime before the full coincidence limit is taken .
it is convenient to have the points take on equal values of @xmath48 and @xmath49 so that the coordinate separation is given by @xmath50 .
the unrenormalised green function now takes on the form : @xmath51 it should be noted at this point that there exists a superficial ultra - violet divergence over @xmath52 and @xmath53 in the above expression .
this divergence can most easily be eliminated using a similar technique as candelas @xcite and anderson @xcite . it is noted that as long as @xmath54 any multiple of @xmath55 can be added to ( [ eq : unreng ] ) .
substituting @xmath43 from ( [ eq : x ] ) in the large @xmath52 and @xmath53 limit and subtracting this term from ( [ eq : unreng ] ) the logarithmic divergences can be eliminated giving the following expression : @xmath56 to calculate the renormalised value of @xmath0 a point splitting technique will be used .
the dewitt generalisation to schwinger s expansion is used as an approximation for the green function .
this term will then be subtracted from ( [ eq : finiteg ] ) and the @xmath57 limit will be taken along the shortest geodesic separating the points .
the dewitt - schwinger counter - term is given by : @xmath58 where @xmath59 is the `` world function '' of synge @xcite which is equal to half of the square of the geodesic distance between two points .
the points in this case will be @xmath60 and @xmath61 .
it can be shown , by geodesic expansion ( see appendix a ) , that the world function in the spacetime considered here takes on the form @xmath62 using these expressions the dewitt - schwinger counter term is equal to @xmath63 the first term in ( [ eq : moreds ] ) can be rewritten in a more convenient way by using the plana sum formula @xcite as used by anderson .
the formula is @xmath64}{e^{2\pi t}-1}.\ ] ] with this , the first expression in the counter term can be written as @xmath65 which gives , for the entire counter - term in the @xmath66 limit : @xmath67 it can be shown that the second and third terms in ( [ eq : fullcounter ] ) , which normally diverge at the horizon , will cancel each other out on the horizon when @xmath24 is equal to the black hole temperature .
that is , when @xmath68 therefore , the renormalised expression for @xmath0 in the hartle - hawking vacuum state is @xmath69 where , for convenience , the following notation has been used @xmath70
in this section the value of @xmath0 will be calculated from ( [ eq : phisq ] ) .
the solution is found by iteratively solving for the function @xmath43 using ( [ eq : xevolve ] ) in the mode functions with the lowest order term defined as : @xmath71 + \frac{9}{4}-\frac{36m^2}{\alpha^6\rho^6}+\frac{\xi r}{\alpha^2}h,\ ] ] and @xmath72 given by @xmath73 with this choice , the lowest order term will be valid at both the horizon and as @xmath31 approaches infinity .
this can also be verified by comparison with ( [ eq : infbehav ] ) and ( [ eq : horizbehav ] ) .
it is most convenient to do the integration over @xmath53 first followed by the sum over @xmath52 and finally , the sum over @xmath74 .
this scheme leads to analytic solutions to both the integrals and the @xmath52 sums . in the solution ,
the sums and integrals have the following form : @xmath75 where @xmath76 is a function of @xmath31 only .
such integrals are known and are given by @xmath77 where @xmath78 is a fractional constant which depends on the particular value of the integer @xmath79 .
the resulting sums over @xmath52 can now be done analytically by the standard contour integration : @xmath80 this is valid as the denominator never becomes singular at integer values of @xmath52 . in the expansion of @xmath0
there appears a term of the form @xmath81 which arises when the lowest order term is combined with the counterterm in ( [ eq : finiteg ] ) and integrated over @xmath53 .
this is the only term which is singular at @xmath82 .
the @xmath82 term is therefore removed from the sum and redefined as to remove a spurious divergence of the form @xmath83 which is present in the same term once the sum from @xmath84 to @xmath85 has been calculated ( how this term arises is shown in more detail in appendix b. ) this redefinition can be performed since the counterterm introduced in ( [ eq : finiteg ] ) may be any quantity and the particular value shown there is exact only in the large @xmath52 limit . the final sum over @xmath74 can now numerically be shown to converge by computing the values of @xmath0 for large @xmath74 .
it can also be shown that the @xmath86 mode makes no contribution to the sum .
the boundary values of the modes must be evaluated before numerical integration of the mode equation can be done .
the value of the mode functions at infinity can easily be seen by studying the mode equation in the asymptotic region and therefore only the value at the horizon is left to be determined . at the horizon , there are many quantities in the expansion of @xmath0 which are inversely proportional to some power of the metric function @xmath87 . by performing an expansion of @xmath0 in the quantity @xmath88 , where @xmath89 is the horizon value of @xmath31 , one can show ( although the procedure is lengthy ) that all terms with @xmath33 raised to some power in the denominator cancel at the horizon . in appendix
b it is also demonstrated how terms which normally make a dominant contribution to @xmath0 cancel here .
it should be noted that the horizon value is directly proportional to the value of the cosmological constant .
this is due to the fact that the spacetime is an einstein spacetime with @xmath90 .
@xmath0 was computed for the conformally coupled , massless case and the result is shown in fig.1 .
it can be seen the the maximum value of @xmath0 occurs near , but not at , the horizon .
this behaviour is analogous to the extreme reissner - nordstrm case @xcite .
this is because , as shown in the appendix and earlier , most contributions to @xmath0 at the horizon vanish .
however , near ( but not on ) the horizon terms with a @xmath91 behaviour make a large contribution . for large @xmath31 most terms in the field expansion vanish and therefore @xmath0 approaches a value which is dominated by the last two terms in ( [ eq : phisq ] ) .
the author would like to thank dr .
k. s. viswanathan for advice and help without which this work would not have been possible .
the author would also like to thank dr .
p. anderson of wake forest university for his patience in answering questions via e - mail .
n. kaloper was also very helpful in drawing my attention to some references which were originally missed .
in this appendix it will be demonstrated how to calculate the world function , @xmath59 , using the method of geodesic expansion .
equations here involve quantities which are defined at different spacetime points . for a brief review of handling bitensors the reader
is referred to christensen @xcite and synge @xcite .
let @xmath92 and @xmath93 be two points in the spacetime close enough together such that they are connected by a _
unique _ geodesic .
the geodesic equation @xmath94 yields the power series @xmath95 where @xmath96 and @xmath97 the last term in equation ( [ eq : geoexp ] ) is zero since the spacetime is static .
( [ eq : geoexp ] ) can be inverted and used in the definition @xcite @xmath98 giving @xmath99 calculating the christoffel symbols using ( [ eq : metric ] ) and noting ( [ eq : xidef ] ) it can be shown that this expression reduces to the one in ( [ eq : sigmas ] ) .
this expression has the same functional form as that of the world function for a static , spherically symmetric spacetime @xcite if the coordinate separation there is also chosen as ( [ eq : xidef ] ) .
this is due to the fact that the @xmath100 vectors eliminate any dependence of @xmath101 on @xmath102 and @xmath103 .
in this appendix the dominant terms of @xmath0 at the horizon will be calculated .
this is useful since the value at the horizon needs to be evaluated as a starting point for the calculation of the mode functions .
this calculation is also useful as it provides insight as to how the @xmath74 counter - term acts to regularise the field . at the horizon ,
the dominant terms in the field expansion are given by @xmath104 for the moment , we choose to ignore the @xmath82 term and concentrate on the first expression in ( [ eq : horizphi ] ) which can trivially be re - written as @xmath105 the product in the above expression is well known yielding the following for ( [ eq : prod ] ) @xmath106 .
\label{eq : nleft}\end{aligned}\ ] ] for very large @xmath74 or ( as in the case here ) very small @xmath72 this becomes @xmath107 .
\label{eq : lzero}\ ] ] if we define the @xmath82 term to cancel out the second term in ( [ eq : nleft ] ) as described in the text , the resultant expression gives @xmath108 which is cancelled by the @xmath74 sum in the counter - term .
this leaves a small constant contribution to @xmath0 at the horizon . | the renormalised value of @xmath0 is calculated for a massless , conformally coupled scalar field in the hartle - hawking vacuum state .
this calculation is a first step towards the calculation of the gravitational back reaction of the field in a black cosmic string spacetime which is asymptotically anti - desitter and possesses a non constant dilaton field .
it is found that the field is divergence free throughout the spacetime and attains its maximum value near the horizon . |
a scattering experiment , using neutrons or x - rays for example , is one of the most direct measures of the structure of materials .
naively , this comes about because in the born approximation , which usually applies for x - rays and neutrons , the intensity in scattering measurements is proportional to the fourier transform of a density - density correlation function .
it is the wavelike properties of the scattering probe which produces the fourier transform . for
a deeper understanding of the relationship between scattering intensity and structure one must realize that this direct correspondence applies precisely only for coherent waves .
indeed , for conventional sources , a given point in the incident wave is only coherent within a small volume of neighboring points .
this coherence volume has transverse dimensions determined by how parallel the wavefronts are and a longitudinal length determined by how monochromatic the wave is . in a standard scattering experiment the different coherence regions of the incident beam scatter independently .
the intensity measured thus depends on an incoherent average over different regions of the scattering volume . by restricting the scattering volume of the sample to less than the coherence volume of the beam
, one can eliminate this incoherent average , and thus learn more about the material s structure .
of course , the experimental difficulty which arises is to obtain sufficient diffracted intensity to measure a signal .
recently , it has been demonstrated that coherent diffraction experiments can be performed with x - rays using high - brilliance synchrotron sources@xcite . for coherent diffraction
, the scattering from an inhomogeneous material displays a characteristic speckled scattering pattern .
for instance , the random distribution of phase - ordering domains shown in fig . 1 and discussed below , results in the speckle pattern shown in fig
. 2 . as the domains change shape , the speckle pattern changes , and this time dependence of the speckle offers a unique method for studying the evolution of inhomogeneous materials . motivated by these advances in experimental techniques ,
we have undertaken a theoretical study of intensity fluctuations caused by scattering from a nonequilibrium system undergoing phase ordering by domain growth . when a disordered homogeneous material is rapidly brought to a new set of conditions , corresponding to the coexistence of two equilibrium phases , a spatial pattern of domains of the two phases develops @xcite .
this change of conditions is often accomplished by a rapid quench from a high temperature to a low one below a miscibility gap .
the quench from a homogeneous state with local fluctuations creates a microstructure of interconnecting , interlocking domains through the kinetics of a first - order phase transition .
as time goes on the domains grow , so as to minimize the area of the domain walls that separate the phases . when the average domain size @xmath9 at time @xmath1 is large compared to all other relevant lengths , except for the extent @xmath10 of the system itself
, the system looks invariant if all lengths are measured in units of that domain size . in this case
the structure of these many domains is said to `` scale '' with @xmath9 .
experimentally , the growing domain structure is often studied by means of the scattering intensity @xcite , whose width is proportional to the inverse of @xmath9 . when the average scattering intensity is scaled in units of this time - dependent length ,
one obtains the scattering function for late times in the form of a time - independent scaling function .
the time - dependence then enters only through @xmath9 , which can typically be described in terms of an exponent @xmath11 , such that @xmath12 .
this scaling hypothesis has been found to apply to a large range of systems , and to be unaffected by many of the microscopic details of specific materials .
that is , the scaling function and the growth exponent @xmath11 are two features which are common to a large number of systems , collectively called a universality class .
universality classes for phase ordering by domain growth are delineated chiefly by the presence or absence of conservation laws .
indeed , many aspects of this nonequilibrium process can be described by relatively simple theoretical models .
for example , for systems described by a nonconserved scalar order parameter , often called model a @xcite , the growth exponent is found to be @xmath13 .
systems included in this class are the ising model with spin - flip dynamics , binary alloys undergoing an order - disorder transition , and some magnetic materials with uniaxial anisotropy .
model b @xcite refers to systems in which the scalar order parameter is conserved , and the only growth mechanism is diffusion .
systems in this universality class have @xmath14 and include the conserved ising model , as well as binary alloys undergoing phase separation . in the present paper
we investigate the time - dependent fluctuations _ around _ this scaling behavior , and we demonstrate how to study these fluctuations experimentally through analysis of the time - dependent scattering .
the first theoretical study of such behavior is given in ref .
the scattering intensity is related to the fourier transform of the order parameter @xmath15 , the scalar field describing the inhomogeneity of a specific sample of the scattering material , by @xmath16 where we ignore the proportionality constant for convenience .
the average of @xmath17 over an ensemble of initial conditions is the structure factor , @xmath18 here , the ensemble average expresses the distinction between coherent scattering , given by @xmath19 , and incoherent scattering , given by @xmath20 .
fluctuations around the average scattering are the main topic of interest in this paper .
the structure factor can also be expressed as the fourier transform of the correlation function of @xmath15 , @xmath21 because of these relationships , the scattering intensity and structure factor have been important tools for studying the dynamics of materials far from equilibrium . if a system is isotropic , then its average scattering properties depend only on the magnitude of the scattering wavevector , @xmath22 .
as mentioned above , scaling by the domain size @xmath12 implies that correlations are time independent when measured in units of @xmath23 .
specifically , for an isotropic system , @xmath24 , where @xmath25 is one form of the scaling function .
fourier transformation gives the average scattering intensity in terms of another form of the scaling function , which depends only on a _ scaled _ time @xmath26 , @xmath27 where @xmath28 is the spatial dimension of the scattering material .
the specific form of @xmath29 depends on the dynamic universality class . as discussed above , when the scattering involves _ coherent _ radiation , @xmath17
is directly measured without self - averaging .
the scattering from an inhomogeneous material displays a characteristic speckled scattering pattern much like the one in fig . 2 , which shows the squared norm of the fourier transform of the configuration in fig . 1
the intensity of each speckle in @xmath30 is due to correlations in the inhomogeneous sample , and @xmath31 gives the fluctuations around the average scattering intensity . for materials in equilibrium , deviations of @xmath19 around the structure factor are induced by thermal fluctuations in @xmath32 .
in fact , fluctuations in the speckle patterns from scattered laser light have been used as the standard basis of photon correlation experiments at wavelengths ranging from the ultraviolet to the infrared @xcite . with the advent of high - brilliance synchrotron photon sources , experiments involving coherent x - rays have become possible .
for example , brownian diffusion rates in gold colloids have been determined from the time required for a change in an x - ray speckle pattern @xcite .
speckle from coherent x - rays has also been used to study equilibrium fluctuations in fe@xmath33al near an order - disorder transition @xcite , and in micellar block - copolymer systems @xcite .
x - rays can probe materials on much smaller length scales than is currently possible with lasers , and their greater penetration allows the study of optically opaque materials .
in the present paper , we present a theoretical study of fluctuations in the scattering intensity from a nonequilibrium system undergoing phase ordering by domain growth . for such systems
the intensity fluctuates around a time - dependent structure factor @xcite .
the time evolution of the intensity at a specific wavevector for one particular quench can be normalized by the average behavior .
a typical example of such a normalized time series , obtained from a simulation at zero temperature , is presented in fig .
it is the correlations of such time series , averaged over many individual speckles and quenches , that can be used to study the pattern formation process in phase - ordering materials . in fig .
3 we also show a `` brownian '' function which was constructed to have the same single - time probability density and an exponential two - time covariance with the same characteristic time as the normalized nonequilibrium scattering intensity . qualitative differences are immediately evident in the two time series .
the brownian function fluctuates quickly , with large amplitude variations , while the intensity fluctuations produced by the phase - ordering system vary slowly , with markedly less variation in amplitude on short time scales @xcite .
this property of the nonequilibrium intensity fluctuations is called persistence @xcite , and it indicates a qualitative difference between the two - time correlation functions for the two processes @xcite . for the remainder of this paper , we specialize to model a as a simple model for systems undergoing phase ordering following a quench through a second - order order disorder phase transition .
some preliminary numerical results from simulations less extensive than the ones used here were presented in ref .
@xcite .
the outline of the rest of this paper is as follows .
section [ sec : meth ] describes the details of our numerical approach , which uses a standard time - dependent ginzburg - landau equation with a nonconserved order parameter . in sec .
[ sec : corr ] the time - time covariance of individual speckle intensities is discussed , using analysis readily adaptable to experimental data . the equality between this covariance and the square of the two - time structure factor of the nonequilibrium material is argued in sec .
[ sec : gauss ] , and comparisons to specific two - time theories follow in sec .
[ sec : theory ] .
there an analytic expression for the universal scaling form for the intensity covariance at late times is obtained .
finally , our results are summarized in sec .
[ sec : sum ] .
the dynamics of speckle in @xmath0-space were simulated by generating successive scattering patterns from a real - space simulation of the dynamics of phase ordering following a quench through an order - disorder phase transition .
the configuration of the real - space system is described by a nonconserved scalar order - parameter field @xmath34 , and its dynamics are governed by the following time - dependent ginzburg - landau equation ,
@xmath35}{\delta \psi({\bf r},\tau ) } + \zeta({\bf r},\tau ) \;. \label{eq : tdgl}\ ] ] the first term on the right - hand side of this langevin equation corresponds to deterministic relaxation , with rate constant @xmath36 , towards a minimum value of the free - energy functional @xmath37 $ ] . thermal noise , which we neglect , is modeled by the random variable @xmath38 , whose intensity is proportional to @xmath36 and temperature by virtue of a fluctuation - dissipation theorem @xcite .
we neglect @xmath38 because the most important sources of noise here are the initial conditions , which give the random domain morphology .
the main effect of thermal noise at low temperatures far below any critical point , is only to thermally roughen the domain walls .
we employ the standard ginzburg - landau - wilson free energy @xcite , @xmath39=\int d{\bf r } \bigg[-\frac{a}{2}\psi^2({\bf r},\tau ) + \frac{u}{4}\psi^4({\bf r},\tau ) + \frac{c}{2}|{\bf\nabla}\psi({\bf r},\tau)|^2 \bigg ] \;.\ ] ] for @xmath40 the local part of the integrand in eq .
( [ potential ] ) represents a bistable potential , and the parameters of the model define an equilibrium field magnitude , @xmath41 , a thermal correlation length , @xmath42 , and a characteristic evolution time , @xmath43 .
the model can be rescaled by choosing a new field @xmath44 , a new length @xmath45 , and a new time @xmath46 . dropping the tilde from the rescaled quantities for convenience , we get the rescaled dynamical equation , @xmath47 which has no adjustable parameters .
since thermal fluctuations are explicitly omitted in the simulation , the only randomness comes from the high - temperature state the system is in before the quench .
this was implemented by an initial condition such that @xmath48 consists of independent random numbers uniformly distributed between @xmath49 .
the simulations were conducted on square lattices with periodic boundary conditions , lattice constant @xmath50 , and a system size of @xmath51 .
the laplacian in eq .
( [ eq : eom ] ) was implemented using the eight - neighbor discretization @xcite @xmath52 where @xmath53 are the four nearest neighbors of @xmath32 ( along the lattice directions ) , and @xmath54 are its four next - nearest neighbors ( diagonally ) .
a simple euler integration scheme with @xmath55 was used to collect data up to a maximum rescaled time of @xmath56 for @xmath57 separate sets of initial conditions .
usually , the order parameter takes the values @xmath58 everywhere except at the domain walls , which are negligibly thin compared to the domains themselves . after our rescaling , domain walls have a soft nonzero width of approximately @xmath59 . to minimize the effect of this nonzero , though small , width
, we use a nonlinear mapping of the rescaled order parameter to @xmath60 before taking the fourier transform .
the transformed field , @xmath61 , is defined by @xmath62 where the fact that the lattice spacing is unity in all directions has been used .
the brillouin zone in two dimensions is defined by the discrete set of wavevectors , @xmath63 , @xmath64 with @xmath65 .
the scattering intensity @xmath66 and its average over initial conditions is the time dependent structure factor , @xmath67 , as described in sec .
[ sec : int ] . to be consistent with the numerical integration , the magnitude of the scattering wavevector , @xmath68 ,
is defined using the operator relation @xmath69 substituting the discrete version of the laplacian from eq .
( [ eq : laplacian ] ) , one obtains for @xmath70 @xmath71 this method has been used to calculate the magnitude of the wavevector for scaling structure factors and data binning . because of lattice effects we consider only those wavevectors with @xmath72 . in order to apply the scaling ansatz
, one must know the characteristic length at time @xmath1 .
often @xmath9 is estimated from the structure factor , however we have found an analytic expression that works well . indeed ,
one advantage of our numerical approach , as compared to monte carlo or cell dynamical simulations , is that we can test theories without using any free parameters .
in particular , ohta , jasnow , and kawasaki @xcite found that the time dependence of the domain size obeyed @xmath73 , with @xmath74 .
furthermore , they found that the structure factor scaled , and they gave an explicit form for the scaling function @xmath29 . in comparison ,
the theory of kawasaki , yalabik , and gunton @xcite gives the same form of @xmath29 , but the factor @xmath75 does not appear in their result for @xmath9 .
our present simulations agree with the amplitude given by ohta _
et al . _ ,
and so we choose a scaled time @xmath76 given by @xmath77 ^ 2 = 4\rho_d k^2 \tau \;.\ ] ]
a quantity to which experiments give ready access is the fluctuation in the speckle intensity as a function of time .
the relationship between an individual speckle at two different times @xmath78 and @xmath79 is , on average , described by the intensity covariance , @xmath80 for random systems , the covariance is maximum in the equal - time limit , @xmath81 , and as the two measurement times become widely separated , the values of the intensity become stochastically independent and the covariance decays to zero . for this relaxational system
, negative values are not expected .
the scaling ansatz extended to this situation allows collapse of the covariance at different @xmath82 by @xmath83 where @xmath84 is the scaling function for the covariance .
the simulated scattering intensities were analyzed in the following way . in eq .
( [ eq : scalcov ] ) , the ensemble average over initial conditions is scaled to the universal form . to make estimates of @xmath85 from the simulations
, we changed the order of the averaging and used the single - simulation averages @xmath86 and @xmath87 .
@xmath86 is the scaled intensity , @xmath88 , averaged over all pairs of @xmath89 that map onto @xmath76 , and @xmath87 is the scaled product of the intensities at two different times , @xmath90 , averaged over all triples @xmath91 that map onto @xmath92 . due to the large amount of data involved in the present study ,
samples for @xmath93 , @xmath94 and @xmath87 were accumulated in a two - dimensional structure of bins organized by a pair of variables related to @xmath92 as described below .
after accumulation into the bins @xmath92 , averaging over independent runs , i.e. , over different initial conditions , was performed to further improve our statistics .
these averages were then used to find the scaling function @xmath84 .
it should be noted that all the quantities used here are also readily obtained experimentally , except for two unknown proportionality constants that depend on details of the experimental system .
one occurs in the equation for the scaled time , corresponding to @xmath95 in eq .
( [ eq : t ] ) .
the other is the proportionality constant between the measured scattering intensity and the squared - norm of the fourier transform of the order parameter , which we have ignored in eq .
( [ eq : sfactor ] ) .
neither of these should be a barrier to comparing our results to experiments .
the normalized analog of the covariance is the correlation function @xcite , @xmath96 @xmath97 is unity by construction , which removes the equal - time variations in the covariance as @xmath1 changes . using the definitions of @xmath98 and @xmath99 above , the scaled version can be expressed as @xmath100 where @xmath101 denotes averaging over initial conditions as before . the contour plot of @xmath102 in the @xmath92-plane , fig . 4 ,
shows how the correlations in individual speckle intensities decay . in this figure
, the line @xmath103 extends along the diagonal .
moving away from that line , contours at values of @xmath104 , @xmath105 , @xmath106 , and @xmath107 are shown .
the scatter in the data is apparent for the last contour and becomes dominant for values of the correlation less than that .
the striking feature of this figure is that the correlations increase in a nontrivial way as the phase - ordering continues .
thus , normalization by the time - dependent intensity is not sufficient to convert the speckle intensity to a stationary time series . a more natural set of variables for studying this effect is @xmath108 and @xmath109 a constant value of @xmath5 corresponds to a line perpendicular to the @xmath110 diagonal , while @xmath111 measures the distance ( in units of scaled time ) away from the diagonal .
the symmetry under exchange of @xmath112 and @xmath113 is retained .
these variables were used for the binning of simulation results that produced fig . 4 , with data grouped into @xmath114 equally wide series for @xmath115 ; each @xmath5 series having @xmath57 equal - width bins with @xmath116 .
the characteristic time difference , @xmath117 , required for the scaled intensity covariance , @xmath84 , to decay to half its maximum value can be found as a function of @xmath5 .
( the normalized correlation function , corr@xmath92 , does _ not _ decay to @xmath118 for small values of @xmath5 . ) here , results for @xmath98 and @xmath99 for @xmath119 ( in @xmath120 series , with @xmath121 bins within each series ) were collected in addition to those previously mentioned .
the value of @xmath117 for each @xmath5 was determined by linear interpolation , and the dependence of @xmath117 on @xmath5 is presented in log - log form in fig . 5 . in this figure ,
two asymptotic limits giving different algebraic relationships are obvious . for small values of @xmath5
the relationship is linear , with the exponent determined by least squares being @xmath122 for @xmath123 . at large values of @xmath124
a least - squares fit for @xmath125 gives an exponent of @xmath126 .
the estimates of error here are from obtaining the exponents from different ranges of @xmath5 in the simulation data ; statistical error is an order of magnitude smaller than these estimates .
the exponents we obtain are in good accord with theories discussed in sec .
[ sec : theory ] below , which give them to be 1 and @xmath118 , respectively .
however , the connection to theory requires some further justification , given in the following section .
while experiments can measure time correlations readily through the covariance or the correlation of intensities , cov or corr , theories to date have made use of the two - point , two - time order - parameter correlation function , @xmath127 . in this section
we argue for , and empirically demonstrate , equality between the intensity covariance for @xmath128 , which involves fourth moments of the order parameter , and the square of the spatial fourier transform of the two - time order - parameter correlation function in our model .
the relationship is exact when @xmath129 is a joint gaussian random number , with its real and imaginary parts independent and gaussian , as we will show below .
however , establishing @xmath130 to be approximately gaussian in the present case is not trivial .
gaussian variables are a natural consequence of the central - limit theorem , which requires a large number of uncorrelated contributions to the variable ( with some restrictions on the properties of the individual contributions ) .
for example , in a disordered system in equilibrium , correlations exist only on the scale of a small length @xmath131 , so a system of edge - length @xmath10 consists of on the order of @xmath132 independent , uncorrelated parts .
then the central - limit theorem applies , and @xmath129 is a complex , gaussian variable . for an ordered system in equilibrium , this argument applies to the fluctuations around the ordered state .
the present nonequilibrium situation has some rough analogies to a disordered equilibrium state . at a given time , the average domain size is @xmath9 , and so the number of independent parts at a given time is approximately @xmath133 , which can be large .
it therefore seems reasonable to expect the central - limit theorem to apply , and indeed , we find empirically that @xmath129 is gaussian .
however , unlike a disordered system , the distribution of domains of different sizes is broad in this case due to the initial long - wavelength instability and , furthermore , it is clear that domains interact as they grow . the degree to which this correlation is important is a nontrivial issue ; below , we test it numerically and find these correlations to be negligible for the two - point quantities of interest in this work . if @xmath129 is gaussian , it is straightforward to relate the intensity covariance @xmath134 to the order - parameter correlation function @xmath135 .
wick s theorem can be used to decompose the intensity - intensity average as @xmath136 here @xmath137 is the two - time structure factor corresponding to the two - point , two - time order - parameter correlation function @xmath138 for @xmath8 , eq .
( [ eq : correspond ] ) can be rewritten as @xmath139 which equates the speckle intensity covariance with the square of the two - time structure factor of the system .
finally , the scaling ansatz defines a universal two - time form @xmath140 and @xmath141 our numerical tests show that this equality holds well in these simulations for @xmath142 . in computer simulations , unlike scattering experiments , the two - time structure factor can be found directly using eq .
( [ eq : skt1t2 ] ) . generally , the product @xmath143 is a complex number , but the mean value of the imaginary part is zero , so the two - time structure factor is real valued . in our simulation data , the imaginary part of @xmath144 is found to be zero , within our accuracy .
results for the real part at @xmath145 and @xmath146 are presented on a log - log scale in fig .
the direct measurements of the two - time structure factor and the square - root of the intensity covariance agree quite well , except at very large values of @xmath147 , where lattice effects are important .
another consequence of our proposed decoupling , leading to the relationship between @xmath134 and @xmath20 , can be tested by simulation . when @xmath81 , eq .
( [ eq : equality ] ) is simply @xmath148 and scaling can be applied to give @xmath149 .
these are compared in fig . 7 , where the agreement is clearly quite good . as an aside , since the variance does not depend on the system size , eq .
( [ eq : var ] ) demonstrates that @xmath17 is a spatially non - self - averaging quantity @xcite .
because of this , increasing the system size @xmath10 will not improve the estimate of @xmath150 given by a specific number of speckles
. however , this is compensated by the fact that more independent speckles are available in a given range of @xmath0 for each trial .
numerical results obtained by shinozaki and oono in a study of spinodal decomposition in three dimensions using the cell - dynamical method appear to be consistent with this result @xcite .
we also note that the normalization of the two - time intensity correlation function , eq . ( [ eq : corr ] ) , can be simplified using eq .
( [ eq : var ] ) .
equation ( [ eq : var ] ) is a property of exponentially distributed variables , and to the extent that @xmath151 is a gaussian random number , @xmath17 will be an exponentially distributed random number .
that is , the probability density for the _ normalized _ intensities , @xmath152 , should satisfy @xmath153 independent of @xmath89 .
the probability density @xmath154 is normalized and has unit mean and standard deviation .
since @xmath154 is identical for all values of @xmath89 , only one density function needs to be constructed .
the results for all @xmath155 are presented for two times in log - linear form in fig .
the histogram for @xmath156 is constructed with a bin size of @xmath157 and the normalized intensity is found using the circular average of @xmath17 from the same trial .
the lower bound on @xmath147 is chosen such that at least @xmath158 speckles contribute to this circular average .
this histogram is accumulated over all @xmath57 trials and then normalized ; for each histogram on the order of @xmath159 samples are available .
the solid line is the expected density , @xmath160 . for the earlier time the probability density is exponential for all @xmath156 . at the latest simulation time the density is exponential only for @xmath161 , with a higher than expected occurrence of speckles brighter than this .
still , less than @xmath162 of the points lie this far into the tail , and the deviation is inferred to be a finite - size effect because it becomes more pronounced as our simulation continues .
this agrees with earlier simulations @xcite , in which increasing the system size eliminated the deviation .
we believe this effect is an artifact of the periodic boundary conditions which allow stabilization of slab - like domain patterns @xcite , producing `` frozen - in '' interference patterns that remain bright while the average intensity steadily decreases .
no @xmath147 dependence is found for @xmath163
two theoretical predictions for the two - point , two - time order - parameter correlation function exist , which will be described and then compared to our simulation results .
the first is an analytic theory , due to yeung and jasnow @xcite , which is an extension of the analysis by ohta , jasnow and kawasaki @xcite .
the second theory , developed by liu and mazenko @xcite , is numerical .
both theories produce approximations for the two - point , two - time order parameter correlation function , @xmath164 , in the asymptotic limit .
these then give the structure factor through the general relation for the fourier transform of a spherically symmetric function @xmath165 in @xmath28 dimensions , @xmath166 where @xmath167 is a bessel function of the first kind of order @xmath168 . in the scaling regime , where the domain - wall thickness can be ignored , the yeung - jasnow correlation function is @xcite , @xmath169 \;.\ ] ] with the previously defined variables , @xmath3 and @xmath4 , this yields @xmath170 \;,\ ] ] expansion of @xmath171 about @xmath172 , followed by term - wise integration , gives @xmath173 this infinite series is convergent for all physical values of @xmath174 and @xmath5 ; however , its terms are in general nonmonotonic in @xmath175 . only as one or both of the scaled times @xmath112 and @xmath113 approaches zero , is the series well approximated by the @xmath175=0 term . the asymptotic early - time form , @xmath176 is therefore a good approximation to eq .
( [ eq : yjdef ] ) only in the very restricted region , @xmath177 near the @xmath112 and @xmath113 axes or the origin .
a more useful , analytical result is obtained in the limit of large @xmath5 and small @xmath111 .
then , the largest terms in the series occur in a relatively wide range of @xmath175 near @xmath178 . for large @xmath5 , the exponential factors in eq .
( [ eq : yjser ] ) suppress the small-@xmath175 terms .
the series can then be converted to an integral , and the factorials can be approximated by stirling s formula to give @xmath179 the identity @xmath180 is used to obtain the explicitly integrable form , @xmath181 \;,\ ] ] where @xmath182 .
we note that the full yeung - jasnow result , eq .
( [ eq : yjdef ] ) , as well as the early - time approximation , eq .
( [ eq : yjaxis ] ) , depends on @xmath111 only through the scaling combination @xmath183 .
however , in the asymptotic late - time approximation , eq . ( [ eq : yjint2 ] ) , the natural scaling combination is @xmath182 .
as we shall see below , this analytical result is in excellent agreement with our numerical simulations .
equation ( [ eq : yjint2 ] ) is readily integrated to yield the explicit large-@xmath5 asymptotic scaling function , @xmath184 where @xmath185 is a modified bessel function of the second kind .
the right - hand side of this equation depends only on @xmath7 . in this large-@xmath5 limit
, the asymptotic forms of this scaling function with respect to @xmath7 are @xmath186 the first line of this equation is exact for @xmath187 , where it gives the asymptotic porod - tail limit of the ohta - jasnow - kawasaki result for @xmath188 . note that , in the limit @xmath189 , the second line gives a nonzero value , in contrast to the proper result given in eq .
( [ eq : yjaxis ] ) .
when @xmath7 is on the order of @xmath190 , @xmath191 diverges from the small-@xmath7 approximation given in eq .
( [ eq : yjbigtau ] ) .
a nonrigorous scaling argument @xcite suggests that the order - parameter correlation function should depend on wavevector and time through @xmath192 , when @xmath193 .
the scaling variable @xmath7 obtained above is consistent with this since @xmath13 for model a. in fact , the @xmath194 result is obtained in the asymptotic large-@xmath5 limit if one repeats the above calculation for general @xmath11 , considering the yeung - jasnow form for the correlation function , eq .
( [ eq : yjcorr ] ) , simply as an integrable approximation valid for small @xmath195 . the second theory for two - time correlations in model a is due to liu and mazenko @xcite .
it is an extension of a theory developed by mazenko @xcite to predict the universal part of the two - point , one - time scaled order - parameter correlation function @xmath196 .
the heart of the liu - mazenko theory is the scaling ansatz @xmath197 and the partial differential equation @xmath198 in eq .
( [ eq : lm ] ) , @xmath199 is considered a function of the rescaled variables @xmath200 and @xmath201 .
note that these definitions are in terms of the physical quantities , but the connections to the rescaled variables used in this work are straightforward .
we have found numerical solutions in terms of liu and mazenko s units , which we then converted so that all results shown here are in terms of the rescaling given in section [ sec : meth ] . the mazenko theory for one - time correlations @xcite serves to provide input into the liu - mazenko model through the scaled order - parameter correlation function @xmath202=0 ) and the numerically obtained eigenvalue @xmath203 . since the system is assumed isotropic , eq .
( [ eq : lm ] ) can be reduced to a single partial differential equation in terms of a radial distance @xmath204 and the logarithmic time ratio @xmath205 .
the scaled order - parameter correlation function in the liu - mazenko theory is normalized to @xmath206 , which causes the tangent term in eq .
( [ eq : lm ] ) to diverge .
use of the transformation @xmath207 leads to the small - time solution @xmath208 this solution is only needed to avoid numerical difficulties at @xmath172 for the first time increment . aside from this , eq .
( [ eq : lm ] ) can be solved numerically using a finite differencing scheme that is implicit with respect to the derivatives , but evaluates the tangent term explicitly .
using @xmath209 and @xmath210 we have reproduced fig . 1 of ref . @xcite .
in addition , the exponent for the autocorrelation @xmath211 ( which will be discussed later ) is recovered .
the liu - mazenko results can be compared to the theory presented here by taking @xmath5 to be a parameter and noting that @xmath212 .
the two - time structure factor predicted by the liu - mazenko theory , found using the fourier transform in eq .
( [ eq : spherft ] ) , is @xmath213 where the scaling function @xmath214 itself depends on @xmath28 .
the two - point , two - time order - parameter correlation functions predicted by the two theories can be compared directly with our simulation , without adjustable parameters .
both theories describe the data well in some instances , and poorly in others .
our results are presented in fig . 9 for @xmath215 and @xmath216
. data from our simulation are circularly averaged with bins of width one in the rescaled distance units , then averaged over 80 trials .
the agreement between the yeung - jasnow theory and our simulation is quite good for this choice of times , while the liu - mazenko result is only qualitatively correct . for larger separations in time ,
the yeung - jasnow theory does not work as well , most noticeably in predicting the autocorrelation function , @xmath217 which is equivalent to @xmath218 .
fisher and huse @xcite have argued that for @xmath219 the autocorrelation obeys the power - law @xmath220^\lambda\ ] ] for late times .
( note that , since @xmath221 , the variables used in this paper give @xmath222 = - 2 n \tanh^{-1 } \delta t/2\bar t$ ] . )
fisher and huse give physical arguments for @xmath223 .
they point out that @xmath224 for the @xmath225 glauber model and conjecture that @xmath226 for the two - dimensional spin - flip ising model . in the limit @xmath227
the yeung - jasnow theory gives @xmath228 as is easily seen by setting @xmath229 in eq .
( [ eq : yjcorr ] ) .
the liu - mazenko theory yields @xmath230 and @xmath231 for two and three dimensions , respectively @xcite .
cell - dynamical simulations performed by liu and mazenko @xcite gave @xmath232 for @xmath233 a recent experiment @xcite found @xmath234 for a two - dimensional nematic liquid crystal using video techniques .
for our simulations we measured @xmath235 for several values of @xmath236 the results are presented on a log - log scale versus @xmath237 in fig .
[ in this measurement we did not employ the nonlinear transformation @xmath238 .
this accounts for the fact that @xmath239 , but does not otherwise seem to affect the results . ] in the figure , the data appear to support a power - law decay at the latest time , and the exponent found by fitting the @xmath240 points @xmath241 for @xmath242 is @xmath243 , which is in good agreement with the experiment and the simulations by liu and mazenko
. however , the local effective value of @xmath244 , @xmath245 , which is shown vs. @xmath246 in the inset in fig . 10 , does not show a clear convergence to an asymptotic limit , especially for larger values of @xmath78 . here
@xmath247 is obtained as a three - point finite - difference estimate around @xmath79 , but estimates that smooth the data over wider intervals yield similarly irregular results .
the estimates of @xmath244 obtained from our present simulations are thus somewhat uncertain
. however , the yeung - jasnow prediction , @xmath248 , is clearly violated .
the predictions of the two theories for the two - time structure factor are compared to our simulation data in fig .
the liu - mazenko theory is in semi - quantitative agreement with our simulations at small @xmath249 but falls off much more rapidly at large @xmath147 .
in addition the shoulder present both in our simulation result and in the yeung - jasnow theory is absent in the liu - mazenko result .
the yeung - jasnow theory agrees much better with our simulation , although it does not fall off as fast at large @xmath147 , and its overestimation around the shoulder is seen consistently throughout our simulations .
the prediction for the characteristic scaled time separation , @xmath250 , defined in sec .
[ sec : corr ] , for both theories is compared to our simulation results in fig .
all three agree that @xmath251 for small @xmath5 .
least - squares fits give @xmath252 for our simulation , @xmath253 and @xmath254 for yeung - jasnow , and @xmath255 for liu - mazenko . at large @xmath5 , the @xmath256 behavior of the yeung - jasnow approach results naturally from eq .
( [ eq : yjres ] ) .
this equation can be solved numerically to find @xmath257 and @xmath258 in two and three dimensions , respectively .
our simulation results give @xmath259 the value of @xmath260 separating the two scaling behaviors corresponds quantitatively to the value for which the first term no longer dominates the series expansion of the yeung - jasnow result , given by the relation ( [ eq : yjaxis2 ] ) .
the agreement between the yeung - jasnow theory and our simulation is quite remarkable . on the other hand ,
for @xmath5 not small the liu - mazenko theory crosses over to a region where @xmath117 is independent of @xmath5 and then into another linear region ; neither of these relationships is seen in our simulations .
the analytic expression for the universal form of the two - time structure factor deduced from the yeung - jasnow theory for large @xmath5 is given in eq .
( [ eq : yjres ] ) .
it is tested for several values of @xmath5 in fig .
11 , which shows good collapse of the simulation data for @xmath261 in terms of the scaling variable @xmath262 .
the data also agree quite well with the yeung - jasnow scaling function , @xmath263 for @xmath264 .
indeed , the slow quadratic decay of correlations near @xmath265 is another signature of persistence in the phase - ordering system . in contrast , brownian fluctuations give exponential decay from @xmath266 .
the agreement with the yeung - jasnow theory is remarkable since the scaling of the simulation data uses an analytic expression for the characteristic length , and no adjustable parameters are employed .
using both numerical and analytic methods , we have investigated time - time correlations in the scattering intensity for a two - dimensional system undergoing an order - disorder transition . the correlations are found to obey scaling in terms of the variables @xmath267 and @xmath268 in the large @xmath5 limit , the correlation data collapse onto a universal curve which is a function only of @xmath269 we argue for , and establish numerically , an exponential distribution for the scattering intensity , eq .
( [ eq : exppdf ] ) , and equality between the scattering intensity covariance and the square of the two - time structure factor of the order parameter , eq .
( [ eq : equality ] ) , for @xmath270 .
we use this equality to test theories for the two - time structure factor due to yeung and jasnow , and liu and mazenko .
both theories describe the data well in some instances , and poorly in other cases .
the yeung - jasnow theory is very similar to our simulation results , so long as @xmath111 is not too large . for @xmath219 ,
the liu - mazenko theory gives a better estimate for the autocorrelation exponent @xmath244 .
for large @xmath5 , however , the liu - mazenko theory does not show the same scaling as our simulation results , where the yeung - jasnow theory compares quantitatively well .
our numerical simulations indicate that a definitive experimental treatment of time - correlations during an order - disorder transition is possible by intensity - correlation spectrometry of scattering speckle .
analysis of experimental correlation data should be similar to the procedures discussed for the simulation data in sec .
[ sec : corr ] . for non - conserved systems ,
the experimental scaling function should be well approximated by eq .
( [ eq : yjaxis ] ) , for small @xmath147 , with one adjustable parameter for each axis . with a similar adjustable parameter scheme
, the scaling function should be described by eq .
( [ eq : yjres ] ) for data in the porod tail .
finally , we expect that the equality between the intensity covariance and the squared two - time structure factor also occurs in other phase - ordering systems ; in particular , we expect that it occurs for conserved systems .
that would allow the experimental study of , for example , time correlations in binary alloys undergoing phase separation by spinodal decomposition , which are representative of model b. indeed , preliminary numerical work we have done indicates this .
experiments on such systems would be of considerable value .
we would like to acknowledge useful discussions with k. kawasaki , y. oono , m. m. sano , h. tomita , and particularly b. morin and k. r. elder .
a. r. is grateful for hospitality and support at mcgill and kyoto universities .
research at mcgill university was supported by the natural sciences and engineering research council of canada and _ le fonds pour la formation de chercheurs et laide la recherche du qubec_. research at florida state university was supported by the center for materials research and technology and by the supercomputer computations research institute ( under u.s.department of energy contract no .
de - fc05 - 85er25000 ) , and by u.s.national science foundation grants no .
dmr-9315969 and dmr-9634873 .
p. a. r.s stay at kyoto university was supported by the japan foundation s center for global partnership through u. s. national science foundation grant no .
supercomputer time at the u. s. national energy research supercomputer center was made available by the u. s. department of energy .
a comprehensive review is given by j. d. gunton , m. san miguel and p. s. sahni , in _ phase transitions and critical phenomena _
8 , edited by c. domb and j. l. lebowitz ( academic , london , 1983 ) . ideas of scaling in domain growth follow similar ideas in critical dynamics , p. c. hohenberg and b. i. halperin , rev .
mod.phys . * 49 * , 435 ( 1977 ) .
a recent specialized review is given by a. j. bray , adv .
phys . * 43 * , 357 ( 1994 ) .
e. dufresne , ph .
d. thesis ( mcgill university , 1995 ) ; e. dufresne , m. sutton , k. elder , b. morin , m. grant , b. rodricks , g. b. stephenson , g. b. held , c. thompson , s. g. j. mochrie , s. e. nagler , l. e. berman , and r. headrick ( preprint , 1997 ) .
a rough , quantitative measure of the different degrees of persistence can be obtained by application of range / scale ( r / s ) analysis @xcite to the short sample time series shown in fig .
this yields a hurst exponent of approximately 0.54 for the `` brownian '' function ( near the expected value of 1/2 ) , and a significantly larger value of approximately 0.74 for the normalized scattering intensity .
however , the correlation times of these normalized time series are slowly increasing .
thus , the processes are not stationary , and the direct application of r / s analysis is only to be regarded as an approximate way of estimating the degree of persistence .
it is natural to normalize the covariance by the standard deviations .
in dynamic light - scattering , the autocorrelation function is normalized by the average intensity . for random gaussian fluctuations , these are equal .
( [ eq : var ] ) . | the statistical properties of coherent radiation scattered from phase - ordering materials are studied in detail using large - scale computer simulations and analytic arguments .
specifically , we consider a two - dimensional model with a nonconserved , scalar order parameter ( model a ) , quenched through an order - disorder transition into the two - phase regime .
for such systems it is well established that the standard scaling hypothesis applies , consequently the average scattering intensity at wavevector @xmath0 and time @xmath1 is proportional to a scaling function which depends only on a rescaled time , @xmath2 .
we find that the simulated intensities are exponentially distributed , and the time - dependent average is well approximated using a scaling function due to ohta , jasnow , and kawasaki .
considering fluctuations around the average behavior , we find that the covariance of the scattering intensity for a single wavevector at two different times is proportional to a scaling function with natural variables @xmath3 and @xmath4 . in the asymptotic large-@xmath5 limit
this scaling function depends only on @xmath6 . for small values of @xmath7 ,
the scaling function is quadratic , corresponding to highly persistent behavior of the intensity fluctuations .
we empirically establish that the intensity covariance ( for @xmath8 ) equals the square of the spatial fourier transform of the two - time , two - point correlation function of the order parameter .
this connection allows sensitive testing , either experimental or numerical , of existing theories for two - time correlations in systems undergoing order - disorder phase transitions .
comparison between theoretical scaling functions and our numerical results requires no adjustable parameters . |
merging plays a key role in the hierarchical galaxy formation in the cold dark matter ( cdm ) universe @xcite .
numerical simulations based on the cdm scenario demonstrate how the galaxies build up hierarchically ( e.g. , * ? ? ?
* ; * ? ? ?
in addition to the theoretical ground , the observed mass function of galaxies as well as that for cluster of galaxies have evolved with redshifts , suggestive of the merging process over the cosmic time , though some discrepancy between the prediction and the observation exists ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ?
furthermore , some fraction of galaxies at all redshifts show an ongoing merger . for high - z galaxies ,
we witness an asymmetric structure composed of many blobs , indicative of a major merger @xcite . in the near field ,
recent minor merger can be seen as a tidal stream in the halo such as the sagittarius stream in the galaxy @xcite or the giant southern stream in m31 @xcite .
recent studies have suggested that magellanic clouds ( mcs ) are the promising candidates to exhibit a sign of past merger .
first , @xcite have claimed that kinematical difference between h i gas and older stars in the small mc ( smc ) @xcite is an end result of a major merger occurred in the early stage of the smc formation .
secondly , the large mc ( lmc ) is found to possess an extended stellar halo exhibiting a broad metallicity spread , which can be linked with the nature of hierarchical formation through mergers @xcite .
finally , the lmc and smc might have belonged to a larger magellanic group @xcite , which ensures a gas infall onto the mcs owing to the smaller velocity dispersion inside the mc group .
note that such an infall is hard to realize once the mcs become the galaxy satellites like the present time .
chemical features of mcs have been extensively investigated through the obserations for young and old stars belonging to the field as well as star clusters ( see the review by * ? ? ?
* ) , though there are less studies for the smc where its past information can be obtained only through the star clusters .
the merit of clusters as a tool to trace the chemical evolution is that their ages estimated from the color - magnitude diagram are far more precise than those of field stars .
accordingly , how the metallicity evolved with time , i.e. , the age - metallicity relation of star clusters , deserves the first - ranked information for the study on the history of mcs .
the past ravaging event such as a major merger , if occurred in galaxies , will imprint some signature in their age - metallicity relation . in the general scheme on chemical evolution of galaxies ,
the metallicity gradually increases with time .
if we spot a break in this relation for a galaxy , what does this feature tell us about the formation history of the galaxy ?
in this letter , we report that there exists a clear dip in [ fe / h ] in the age - metallicity relation among the star clusters in the smc , and conclude that it is a relic of a major merger occurred at the early stage of smc formation .
its implication for the history of smc in connection with the lmc and the galaxy is also discussed .
filled circles in figure 1 denote the correlation of [ fe / h ] with age for individual star clusters .
here we have compiled the latest data from various references , and the result for 386 field stars that are divided into six age bins is also shown ( open circles ) . a fairly good agreement between clusters and field stars
is seen . as an overall feature ,
we see a gradually increasing trend of [ fe / h ] with time . at the same time , a gap in level between the evolutionary path for @xmath58 gyr and that for @xmath67 gyr is conspicuous . in other words ,
the fe abundance once decreases @xmath08 gyr ago and then restarts to increase around @xmath7=7 gyr till the present .
here we want to stress that this dip is not an artifact resulting from the systematic difference between the multiple studies . except for one cluster ( at @xmath7=3.5 gyr ,
* ) , ages and metallicities of other clusters are the results of two groups by glatt s et al . and piatti et al .
, each of which occupies the separate age - range in the plot , i.e. , for old ( @xmath8 6 gyr ) clusters , and young and intermediate - age clusters , respectively .
thus , the observed dip is established by one study . moreover , the previous results on old clusters are also suggestive of the presence of dip .
figure 2 shows the data of glatt s group extracted from fig . 1 for the clusters with their ages older than 5 gyr ( black circles ) , together with the results by @xcite ( red circles ) and by @xcite ( blue circles ) .
we see the dip convincingly for one while marginally for the other .
meanwhile , this figure makes the inhomogeneity of metallicity values around @xmath9 - 6.5 gyr by glatt s group stand out
. the lower value of [ fe / h ] for one cluster seems compatible with other two studies . finally , it should be of note that the dip feature is supported by the data of field stars as well as by an increasing rate of [ fe / h ] for young and intermediate - age clusters as shown in 3 . only a mechanism to make such a dip in [ fe / h ] is considered to be an external dilution of the interstellar matter ( ism ) by an accretion of very low - metallicity gas .
the degree of dilution is not small : the decrease in [ fe / h ] is estimated to be 0.3 dex at least .
that is equivalent to an accretion of pristine gas with an equal mass to the existing ism at that time , likely an order of @xmath10 , as the least mass scale .
therefore , the observed decrease in [ fe / h ] in the smc has a strong implication for the past occurrence of a major ( not minor ) merger . in the following ,
we construct the model for chemical evolution of the smc , based on the major merger scenario .
gyr ) clusters by three different groups .
black circles represent the data of glatt s group extracted from fig . 1 for the clusters with their ages older than 5 gyr .
red and blue circles are taken from @xcite and @xcite , respectively . the overall trend implied by the age - binned averages ( crosses ; @xmath11 7 gyr ,
7 @xmath12 10 gyr , @xmath8 10 gyr ) is shown by the arrows . here
we neglect three data situated around ( @xmath7 , [ fe / h])=(6 , -1.1 ) ( see the text).,width=226 ] our prime aim is to reproduce the age - metallicity relation of the smc , by incorporating the major merger phenomenon into the model .
let @xmath13 be the star formation rate and @xmath14 , @xmath15 be the gas infall rate , the accretion rate by a major merger , respectively , then the gas fraction @xmath16 and the abundance of heavy - element @xmath17 @xmath18 in the gas change with time according to @xmath19 @xmath20 where @xmath21 is the mass fraction locked up in dead stellar remnant and long - lived stars , @xmath22 is the heavy - element yield from an snii or sn ia , and @xmath23 ( @xmath24 ) is the abundance of heavy element contained in the infalling gas ( the accreting gas driven by a merger ) . the star formation rate @xmath13 is assumed to be proportional to the gas fraction with a constant rate coefficient of 0.03 gyr@xmath3 . for the infall rate , we adopt the formula that is proportional to @xmath25 with a timescale of infall @xmath26 .
in this formula , @xmath26= 0.3 gyr is adopted , and the metallicity @xmath23 of an infall is assumed to be [ fe / h]=-2 with a sn - ii like enhanced [ @xmath21/fe ] ratio . for reference , the values for the corresponding rate and timescale in the solar neighborhood are predicted to be 0.4 gyr@xmath3 and 5 gyr , respectively @xcite .
the effect of a merger on chemical evolution is devised to be an additional term @xmath15 which has the same form as @xmath14 . here
, the onset of @xmath15 is set at @xmath27 gyr with the assumption that the age of smc is 13 gyr old .
thus , @xmath28 .
the timescale of accretion is assumed to be @xmath26= 1 gyr .
this somewhat long timescale is adopted to avoid the sudden decrease in [ fe / h ] by a merger .
the total gas accretion is normalized so that @xmath29 .
the mass ratio between two functions is the crucial factor to control the impact of the merging .
we calculate three cases of @xmath30 . in all cases ,
the metallicity @xmath31 is set to be [ fe / h]=-2 so as to produce the dip in [ fe / h ] by a merger event . for the initial mass function ,
we assume a power - law mass spectrum with a slope of -1.35 , which is combined with the nucleosynthesis yields of sne ia and ii taken from @xcite to deduce @xmath32 and @xmath33 .
we apply an instantaneous recycling approximation to sn ii progenitors , leading to the ejection rate of heavy elements , which is proportional to @xmath13 .
in contrast , the effect of time delay yields the ejection rate from sn ia s , which is proportional to @xmath34 . it is assumed that @xmath35 spans over some range according to a distribution function @xmath36 .
here we adopt that the fraction of the stars that eventually produce sne ia for @xmath37is 0.05 with a box - shaped @xmath36 for @xmath38 gyr @xcite .
, and are normalized to coincide with the total number of the sample stars of observation . , width=272 ] first , we show the calculated age - metallicity relations for three cases ( dotted curve : no merger , dashed curve : one - to - one merger , solid curve : one - to - four merger ) , together with the observations . here
ages represent the lookback time .
the merger models can reproduce not only the observed dip around 7 gyr ago but also the increasing rate of [ fe / h ] before and after the dip .
there are several clusters deviating from the evolutionary paths predicted by the merger models .
these are classified into two groups ; one is well fitted by the model with no merger ( two clusters ) and the other is composed of ones dropping below the curves due to their lower metallicities than the predictions at each age ( the assemblage of four clusters at @xmath39 gyr and one at @xmath40 gyr ) .
the presence of such clusters might imply that there existed regions where a gas accretion driven by a major merger did not take place , and that additional accretion events , possibly by minor mergers , locally diluted the ism in the smc , along with another possibility that low - metallicity clusters came into the smc from other galaxies through major / minor mergers .
the successful merger models should be verified by checking whether they are compatible with other chemical features of smc . in general , stellar abundance distribution function ( adf ) gives a stringent constraint on the models . for the smc , from over 350 field stars @xcite ,
the observed adf can be constructed .
the distribution denoted by crosses in figure 2 shows the derived adf form their data , in which we exclude stars in three outer fields where the metallicity on average is distinctly lower than other fields . in the end , 293 stars are entered .
the adf thus obtained is compared with the results by the one - to - four merging model ( filled circles ) and the one - to - one model ( open circles ) . a fairly good agreement with the observation , especially for the one - to - four model , can be seen . with [ fe / h ] for smc stars , together with the observed data at present ( filled circle : venn 1999 , open circle : rolleston et al .
solid and dashed curves are the same as for fig .
1.,width=264 ] finally , we show the resultant correlation of [ mg / fe ] with [ fe / h ] for two merging models . after drawing a loop which is a result of the increase - decrease - increase chain of [ fe / h ] induced by a merger ,
both curves reach the present - day observed value . for the present value ,
chemical compositions for a - type supergiant @xcite and b - type dwarf @xcite are shown .
our merging models predict a relatively large scatter in [ mg / fe ] for the range of -1.4 - 1.1 as a reflection of a looping feature .
motivated by the observed dip in [ fe / h ] seen in the age - metallicity relation for the star clusters in the smc , we construct the chemical evolution model in which the gas infall driven by a major merger is incorporated .
accordingly , we claim that the smc experienced a major merger in the manner that two progenitor galaxies of the smc merged with the mass ratio of 1:1 to 1:4 around 7.5 gyr ago . different speed of chemical enrichment between two galaxies results in creating the gap / discontinuity of age-[fe / h ] relation as observed .
@xcite found that the smc formed relatively few stars between 8.4 and 3 gyr ago and thus did not claim the presence of a starburst population around 7.5 gyr ago .
this together with our finding implies that a major merger occurred in the smc did not trigger a major starburst for some physical reasons but proceed with a moderate star formation .
we do not doubt that the detailed processes involved in smc evolution , including the connection between the potential merger events and the star formation history , will be unraveled by the future databases of detailed elemental abundances of old stellar population in the smc , which will be brought by the new - generation of surveys such as hermes . here
we have proposed a galaxy - galaxy merger in the framework that the smc merged with a gas - rich dwarf galaxy including a stellar component , since such process would change the dynamical properties of the smc , the signature of which is suggested to be imprinted in the kinematical difference between h i gas and old stars @xcite . if we put aside the kinematical discussion , an accretion event of a massive hi gas cloud with a mass of @xmath10 order , which is not associated with a galaxy embedded in a dark matter halo ,
also explains the observed [ fe / h ] dip in the smc .
however , such a massive gas accretion event is almost outside the range of possibility
. that s firstly because such massive isolated gas clouds are extremely rare and usually associated with tidal debris as a result of interacting galaxies ( e.g. , * ? ? ?
nowhere in the local group has possessed a potentiality to provide such a huge amount of gas with a short timescale of @xmath41gyr . on the other hand ,
if the smc was already within the galactic halo about 7.5 gyr ago , a gas accretion onto the smc rather than onto the galaxy would be severely suppressed owing to the large relative - velocity between the smc and the cloud in the galactic halo . in the end , it is concluded that the observed dip is attributable to the gas infall associated with a galaxy - galaxy merger occurred to the smc .
the difficulty of a gas accretion onto the smc inside the galaxy discussed above also holds for a merger case .
given that the observed velocity dispersion ( @xmath2 km s@xmath3 , * ? ? ?
* ) of the smc is much smaller than that ( @xmath4 km s@xmath3 estimated from @xmath42 , * ? ? ?
* ) of the galactic halo , the major merger is highly unlikely to happen within the galactic halo .
accordingly , we suggest the following two possible schemes on the smc s merger event .
one is that the merger event could have happened in a small group where ( i ) the velocity dispersion was comparable to that of the smc and ( ii ) there was a number of gas - rich dwarfs .
recent cosmological simulations have shown that the galaxy grows through accretion of small groups ( e.g. , * ? ? ?
* ) , which is compatible with the view that the small group including the smc might have existed before and accreted into the galaxy potential to lose other member galaxies in the galactic halo .
the other is that the smc had been a binary system of two gas - rich dwarfs until @xmath07.5 gyr ago and then merged with each other to form a single dwarf with an extended hi gas disk . by incorporating the hypothesis that the mcs have a common diffuse dark halo with the mass larger than @xmath43 into the orbital models for the mcs
, @xcite succeeded in reproducing the present 3d velocities of mcs , which are consistent with the latest proper motion measurements @xcite . with the same model
, he found that the strong tidal interactions among the lmc , the smc , and the galaxy for the last 2 gyr led to the formation of the magellanic stream .
it is , however , unclear how the possible common halo was formed in the histories of the lmc and the smc . as the possible mechanism to form the common halo
, it can be regarded as the remnant structure evolved from a group where both the lmc and the smc had belonged until recently . this scenario leads to a plausible picture that such a group provided the smc with the major merger event as proposed in the present study .
we note that the recent dynamical coupling of the lmc and the smc around 3 - 4 gyr ago is another promising mechanism for the formation of the common halo in terms of the origin of `` the age gap '' in the globular cluster system of the lmc @xcite .
interestingly , we see a similar dip in [ fe / h ] at @xmath08 - 9 gyr ago for field stars in the lmc @xcite , though in contrast to the continuous age - distribution of field stars , the lmc star clusters show the lack of the intermediate - age ( 4 @xmath44 9 gyr ) clusters ( e.g. , * ? ? ?
* ; * ? ? ?
whether this observed dip in [ fe / h ] for the lmc has some connection with that for the smc at @xmath07.5 gyr ago is the issue to be solved with a high priority .
if the lmc s dip is due to a merger event occurred at the slightly older age than the smc , the event will imprint some observable records in stellar structure and kinematics of the present - day lmc .
the potential candidates are considered to be the observed extended stellar halo ( e.g. , * ? ? ?
* ) and the presence of the thick disk ( e.g. , * ? ? ?
* ) . in the end
, we can bear the following picture in mind .
a small group , so - called the magellanic group , contained both the lmc and the smc at least until the completion of a merging process about 6.5 gyr ago , and then it had been partly or completely disrupted by the galactic tidal field during its accretion onto the galaxy .
it results in the present situation that the other group member galaxies exist as satellite galaxies with orbits similar to the mcs in the local group @xcite .
validation of this scheme together with the issue of which satellite galaxies originally belonged to the magellanic group will be discussed in our future papers .
the authors thank the anonymous referee for useful comments that helped improve this paper .
this work is assisted in part by grant - in - aid for scientific research ( 21540246 ) of the japanese ministry of education , culture , sports , science and technology .
abadi , m. g. , navarro , j. f. , steinmetz , m. , & eke , v. r. 2003 , apj , 591 , 499 bahcall , n. a. , & bode , p. 2003 , apj , 588 , l1 bekki , k. 2008 , apj , 684 , l87 bekki , k. , & chiba , m. 2008 , apj , 679 , l89 bekki , k. , couch , w. j. , beasley , m. a. , forbes , d. a. , chiba , m. , da costa , g. s. 2004 , apj , 610 , 93 binney , j. , & tremaine , s. 1987 , galactic dynamics ( princeton : princeton univ . press ) carrera , r. , gallart , c. , aparicio , a. , costa , e. , mndez , r. a. , & nol , n. e. d. 2008a , aj , 136 , 1039 carrera , r. , gallart , c. , hardy , e , aparicio , a. , & zinn , r. 2008b , aj , 135 , 836 conselice , c. j. , bershady , m. a. , dickinson , m. , & papovich , c. 2003 , aj , 126 , 1183 da costa , g. s. , & hatzidimitriou , d. 1998 , aj , 115 , 1934 dirsch , b. , richtler , t. , gieren , w. p. , & hilker , m. 2000 , a&a , 360 , 133 donghia , e. , & lake , g. 2008 , apj , 686 , l61 glatt , k. et al .
2008a , aj , 135 , 1106 glatt , k. et al .
2008b , aj , 136 , 1703 harris , j. , & zaritsky , d. 2004 , aj , 127 , 1531 harris , j. , & zaritsky , d. 2006 , aj , 131 , 2514 hill , v. 2004 , in carnegie observatories astrophysics series , vol . 4 : origin and evolution of the elements , ed .
a. mcwilliam & m. rauch ( cambridge : cambridge univ .
press ) , p. 203 ibata , r. a. , gilmore , g. , & irwin , m. j. 1994 , nature , 370 , 194 ibata , r. a. et al .
2001 , nature , 412 , 49 kallivayalil , n. , van der marel , r. p. , & alcock , c. 2006 , apj , 652 , 1213 kayser , a. et al .
2007 , in stellar populations as building blocks of galaxies , iau symp . 241 , ed .
a. vazdekis & r. f. peletier ( cambridge : cambridge univ .
press ) , p. 351 koribalski , b. s. et al .
2004 , aj , 128 , 16 li , y .- s . , & helmi , a. 2008 , mnras , 385 , 1365 majewski , s. r. , nidever , d. l. , muoz , r. r. , patterson , r. j. , kunkel , w. e. , & carlin , j. l. 2009 , in the magellanic system : stars , gas , and galaxies , iau symp . 256 , ed . j. th .
van loon & j. m. oliveira ( cambridge : cambridge univ . press ) , p. 51 marchesini , d. et al .
2008 , submitted to apj mighell , k. , sarajedini , a. , & french , r. s. 1998 , aj , 116 , 2395 olszewski , e. w. , schommer , r. a. , suntzeff , n. b. , & harris , h. c. 1991 , aj , 101 , 515 piatti , a. e. , santos , j. f. c. jr , clari , j. j. , bica , e. , sarajedini , a. , & geisler , d. 2001 , mnras , 325 , 792 piatti , a. e. , sarajedini , a. , geisler , d. , seguel , j. , & clark , d. 2005 , mnras , 358 , 1215 piatti , a. e. , sarajedini , a. , geisler , d. , gallart , c. , & wischnjewsky , m. 2007a , mnras , 381 , l84 piatti , a. e. , sarajedini , a. , geisler , d. , gallart , c. , & wischnjewsky , m. 2007b , mnras , 382 , 1203 pozzetti , l. et al .
2007 , a&a , 474 , 443 rolleston , w. r. j. , venn , k. , tolstoy , e. , & dufton , p. l. 2003 , a&a , 400 , 21 springel , v. et al .
2005 , nature , 435 , 629 stanimirovi , s. , staveley - smith , l. , & jones , p. a. 2004 , apj , 604 , 176 tsujimoto , t. , nomoto , k. , yoshii , y. , hashimoto , m. , yanagida , s. , & thielemann , f .- k .
1995 , mnras , 277 , 945 tsujimoto , t. 2007 , apj , 665 , l115 van der marel , r. p. , alves , d. r. , hardy , e. , & suntzeff , n. b. 2002 , aj , 124 , 2639 venn , k. a. 1999 , apj , 518 , 405 yoshii , y. , tsujimoto , t. , & nomoto , k. 1996 , apj , 462 , 266 white , s. d. m. , & rees , m. j. 1978 , mnras , 183 , 341 white , s. d. m. , & frenk , c. s. 1991 , apj , 379 , 52 | the formation history of the small magellanic cloud ( smc ) is unraveled based on the results of our new chemical evolution models constructed for the smc , highlighting the observed anomaly in the age - metallicity relation for star clusters in the smc .
we first propose that evidence of a major merger is imprinted in the age - metallicity relation as a dip in [ fe / h ] .
our models predict that the major merger with a mass ratio of 1:1 to 1:4 occurred at @xmath07.5 gyr ago , with a good reproduction of the abundance distribution function of field stars in the smc . furthermore , our models predict a relatively large scatter in [ mg / fe ] for @xmath1 } -1.1 $ ] as a reflection of a looping feature resulting from the temporally inverse progress of chemical enrichment , which can be tested against future observational results .
given that the observed velocity dispersion ( @xmath2 km s@xmath3 ) of the smc is much smaller than that ( @xmath4 km s@xmath3 ) of the galactic halo , our finding strongly implies that the predicted merger event happened in a small group environment that was far from the galaxy and contained a number of small gas - rich dwarfs comparable to the smc .
this theoretical view is extensively discussed in the framework that considers a connection with the formation history of the large magellanic cloud . |
let @xmath9 and @xmath10 be fixed positive integers such that @xmath11 and @xmath12 .
dirichlet proved in 1837 that the integer @xmath13 is prime infinitely often ; this is dirichlet s theorem on arithmetic progressions .
the prime number theorem tells us that the total number of primes @xmath14 is asymptotic to @xmath15 , i.e. , @xmath16 .
moreover , the generalized riemann hypothesis implies that primes are distributed approximately equally among the @xmath17 residue classes modulo @xmath9 corresponding to specific values of @xmath10 .
thus each residue class contains a positive proportion , about @xmath18 , of all primes below @xmath19 .
accordingly , the average prime gap below @xmath19 is about @xmath20 , while the average gap between primes @xmath7 is about @xmath21 . in this paper
we empirically study the growth and distribution of the values of the function @xmath1 , the record ( maximal ) gap between primes of the form @xmath13 below @xmath19 , for @xmath22 . the special case @xmath23 , i.e. the maximal gaps between primes below @xmath19 , has been studied by many authors @xcite .
computational experiments investigating the actual distribution of ( properly rescaled ) values @xmath1 are of interest , in part , because in cramr s _ probabilistic model of primes _ @xcite there exists a limiting distribution of maximal prime gaps @xcite , namely , the gumbel extreme value distribution .
the fact that values of certain arithmetic functions have limiting distributions is among the most beautiful results in number theory .
a well - known example is the limiting distribution of @xmath24 , the number of distinct prime factors of @xmath25 .
the _ erds - kac theorem _ states that , roughly speaking , the values of @xmath24 for @xmath26 follow the _ normal distribution _ with the mean @xmath27 and standard deviation @xmath28 , as @xmath29 .
note that it is virtually impossible to observe the erds - kac normal curve for @xmath24 in a computational experiment because such an experiment would involve factoring a lot of gigantic values of @xmath25 .
by contrast , computations described in section [ numresults ] and _ appendix _ are quite manageable and do allow one to draw maximal gap histograms , which turn out to closely fit the gumbel distribution .
these results , together with @xcite , support the hypothesis that there is a limiting distribution of ( properly rescaled ) values of @xmath1 . however , a formal proof or disproof of existence of a gumbel limit law for maximal gaps between primes in residue classes seems beyond reach .
[ cols= " < , < " , ] let us begin with a simple example . for @xmath30 , @xmath31
, running the program maxgap.gp produces the results shown in table 1 .
it is easy to check that all gaps @xmath32 in the table satisfy the inequality @xmath33 , suggesting several possible generalizations of _ cramr s conjecture _ ( see sect.[gencram ] ) .
figure [ fig:1 ] shows all values of @xmath32 for @xmath30 , @xmath34 $ ] , @xmath12 , @xmath35 .
the horizontal axis @xmath36 reflects the actual end - of - gap prime @xmath3 of each maximal gap .
the results for other values of @xmath9 closely resemble fig.[fig:1 ] . between primes @xmath37 in residue classes mod @xmath9 , with @xmath30 , @xmath38 , @xmath12 .
plotted ( bottom to top ) : _ average gaps _ @xmath39 between primes @xmath40 in residue classes mod @xmath9 ( solid curve ) ; _ trend curve _ @xmath41 of eq.([trend ] ) ( white dotted curve ) ; the conjectural ( a.s . )
upper bound for @xmath32 : @xmath42 ( dashed line ) .
, width=544,height=393 ] the vast majority of record gaps @xmath1 are near a smooth trend curve @xmath41 : @xmath43 where @xmath44 is the _ expected average gap _ between primes in a residue class mod @xmath9 , defined as @xmath45 and @xmath46 is a correction term .
clearly , for @xmath47 or for any negative @xmath48 , we have @xmath49 we can heuristically derive equations ( [ trend ] ) and ( [ defa ] ) as follows . _ extreme value theory _ predicts that , for @xmath50 consecutive events occurring at i.i.d .
random intervals with cdf @xmath51 ( i.e. at average intervals @xmath52 ) , the most probable maximal interval between events is about @xmath53 , while the width of extreme value distribution is @xmath54 ; see @xcite . as @xmath55 , the distribution of maximal intervals approaches the gumbel distribution @xmath56 with scale @xmath52 and mode @xmath57 . in eqs.([trend ] ) , ( [ defa ] ) we simply take @xmath58 and @xmath59 .
however , extreme value theory alone can not accurately predict the actual behavior of the correction term @xmath46 in ( [ trend ] ) .
it is reasonable to expect that , similar to @xmath44 , the term @xmath46 is a function of both @xmath9 and @xmath60 ( the average prime gap below @xmath19 ) .
indeed , empirically did * _ not _ * work in @xmath1 trend ( [ trend ] ) .
but if instead of @xmath1 we look at the trend of record gaps in an increasing sequence of numbers @xmath25 at _ random intervals
_ @xmath61 , where @xmath62 is a random variable with cdf @xmath63 , then @xmath64 in the trend equation does work .
( we have verified this in computational experiments using the program randomgap.gp ; see _
appendix_. ) ] we have @xmath65 where the approximate values and @xmath66 in terms of @xmath67 , @xmath68 or combinations thereof .
( for instance , could eq.([defc ] ) take a more precise form @xmath69 or possibly @xmath70 ? )
however , the numerical values of @xmath71 and @xmath66 are perhaps not as interesting as the _ unbounded growth _ of the correction term @xmath72 in ( [ trend ] ) .
the situation here resembles legendre s approximation to the prime counting function , @xmath73 ( * ? ? ?
* a228211 ) , where the numerical value of the constant 1.08366 is of historical interest only
. nevertheless , for the purpose of distribution fitting in section [ distsect ] we do need _ some _ realistic formula for the @xmath1 trend , and eqs.([trend])([defc ] ) do this job well , at least for @xmath74 , @xmath75 .
] of @xmath71 and @xmath66 are @xmath76 computations show that formulas ( [ trend])([defc ] ) are applicable at least for @xmath74 , @xmath75 .
( for smaller @xmath9 the data become scarce . ) as we can see in figure [ fig:1 ] , these formulas reflect the actual trend of @xmath1 quite well . equations ( [ defb])([defc ] ) mean that the correction term @xmath77 is unbounded and ( eventually ) negative : @xmath78 as @xmath29 , @xmath79 . combining ( [ defb])([defc ] ) with the inequality @xmath80 for @xmath81 , we can also conclude that @xmath82 we have thus found that the maximal gaps between primes in each residue class are mainly observed within a strip of increasing width @xmath83 around the trend curve @xmath84 of eq.([trend ] ) , where @xmath85 is the expected average gap between primes in the respective residue class . now let us take a closer look at the distribution of maximal gaps in the neighborhood of this trend curve .
we perform a rescaling transformation similar to ( * ? ? ?
* sect.5.2 ) : subtract the trend @xmath84 from actual gap sizes , and then divide the result by the `` natural unit '' @xmath44 .
all record gap values @xmath32 are mapped to _
standardized values _
@xmath86 : @xmath87 figure [ fig:2 ] shows the histograms of standardized maximal gaps for @xmath88 .
( histograms for other @xmath9 look similar to fig.[fig:2 ] . )
we can see at once that the histograms and fitting distributions are skewed : the right tail is longer and heavier .
this skewness is a well - known feature of extreme value distributions . among all two - parameter distributions supplied by the distribution fitting software @xcite
, the best fit is the _
gumbel distribution_. this opens up the question whether the gumbel distribution is the limit law for properly rescaled sequences of the @xmath1 values as @xmath29 ; cf.@xcite .
does such a limiting distribution exist at all ?
if we look at three - parameter distributions , then one of the best fits is the _ generalized extreme value _ ( gev ) distribution , which includes the gumbel distribution as a special case .
the _ shape parameter _ in the best - fit gev distributions is very close to zero ; note that the gumbel distribution is a gev distribution whose shape parameter is exactly zero . and fitted gumbel distributions ( pdf ) for 8- , 9- , 10- , 11- , and 12-digit primes @xmath3 , with @xmath88 , @xmath11 .
, width=544,height=748 ] there are notable differences between our situation shown in fig.[fig:2 ] and the distributions of maximal gaps between _ prime @xmath89-tuples _ shown in ( * ? ? ?
* fig.4 ) even though in either case the gumbel distribution is a good fit , with the majority of record gaps occurring within @xmath90 of the respective trend curve . * in our case ( figs.[fig:1 ] and [ fig:2 ] ) , the available data is not scarce quite the opposite : there are thousands of data points available . on the other hand , data on maximal gaps between prime @xmath89-tuples is scarce ; at present we have fewer than 100 data points for any given type of prime @xmath89-tuple ( * ? ?
* tables 24 ) . * in case of maximal gaps between prime @xmath89-tuples , @xmath91 , a smaller correction term @xmath92 usually works well ; see ( * ? ? ?
* sect.4 - 5 ) .
( in our randomgap.gp model runs , even @xmath93 did work ; see _
appendix_. ) by contrast , for maximal gaps between primes in a residue class , the trend equations ( [ trend])([defc ] ) involve an unbounded correction term @xmath94 .
as noted by brent @xcite , primes seem to be less random than twin primes .
we can add that , likewise , record gaps between primes in a residue class seem to be somewhat less random than those for prime @xmath89-tuples .
our record gaps analysis also shows that primes @xmath3 do not go quite as far from each other as in the randomgap.gp model .
pintz @xcite discusses various other aspects of the `` random '' and not - so - random behavior of primes .
our extensive computations and eqs.([trend])([defc ] ) suggest the following interesting conjecture describing the growth of maximal gaps between primes in residue classes . * conjecture on the trend of * @xmath1 . for any @xmath0 with @xmath12 , there exist real numbers @xmath95 and @xmath96 such that @xmath97 and the difference @xmath98 changes its sign infinitely often . on the other hand
, marek wolf @xcite gives the following approximation for @xmath99 , the _ maximal prime gap below _
@xmath19 , in terms of the prime counting function @xmath100 .
* wolf s conjecture : * @xmath101{wolf2011}}.\ ] ] wolf s reasoning also implies that the difference @xmath102 changes its sign infinitely often .
conjectures ( [ trendconj ] ) and ( [ wolfconj ] ) are related .
indeed , for @xmath103 equation ( [ defc ] ) gives @xmath104 ; therefore , using the fact that @xmath105 and applying the approximations @xmath106 we can rewrite both ( [ trendconj ] ) and ( [ wolfconj ] ) in this common form : @xmath107 so the maximal prime gap predictions ( [ trendconj ] ) and ( [ wolfconj ] ) agree up to @xmath108 .
moreover , one can ensure that ( [ trendconj ] ) and ( [ wolfconj ] ) agree up to @xmath109 using an appropriate choice of @xmath71 for @xmath103 .
in this section we discuss several generalizations of familiar conjectures that are related to ( and in some cases suggested by ) the numerical results of section [ numresults ] .
this discussion places our computational experiments in a broader context .
some of the conjectures ( sect.[gencram ] ) have been proposed by the author earlier on the _ primepuzzles.net _
website @xcite .
we always assume that @xmath110 and @xmath12 .
the riemann hypothesis is a statement about non - trivial zeros of the riemann @xmath111-function .
an `` elementary '' reformulation of the rh is the _ prime number theorem _ with an @xmath112 error term .
it states that @xmath100 , the total number of primes @xmath14 , satisfies @xmath113 the generalized riemann hypothesis is a similar statement concerning non - trivial zeros of dirichlet s @xmath114-functions .
an `` elementary '' reformulation of the grh says that @xmath115 , the total number of primes @xmath7 , @xmath116 , satisfies @xmath117 roughly speaking , the grh means that for large @xmath19 the numbers @xmath115 and @xmath118 almost agree in the left half of their digits .
thus the grh justifies our assumption that average gaps between primes @xmath14 in each residue class mod @xmath9 are nearly the same size , @xmath119 .
consequently , it is reasonable to expect that maximal gaps in each residue class have about the same growth trend and obey approximately the same distribution around their trend .
this is indeed observed in our computational experiments .
( a weaker unconditional result similar to the grh is the siegel walfisz theorem . ) if @xmath99 is the maximal gap between _ primes below _
@xmath19 , cramr @xcite conjectured in the 1930s that @xmath120 .
clearly @xmath121 for all @xmath122 .
we give four conjectures ( from most to least plausible ) , each of which can be considered a generalization of cramr s conjecture to maximal gaps @xmath32 between primes in a residue class modulo @xmath9 .
everywhere we assume that the maximal gap @xmath32 ends at the prime @xmath123 . *
generalized cramr conjecture a. * for any @xmath0 with @xmath12 , we have @xmath124 this weakest generalization of cramr s conjecture seems quite likely to be true .
* generalized cramr conjecture b. * _ almost all _ maximal gaps @xmath32 satisfy @xmath125 for @xmath103 and @xmath31 , the above two generalizations of cramr
s conjecture are compatible with the heuristics of granville @xcite . on the other hand ,
the stronger formulations in subsections [ limsupcram ] and [ naivecram ] contradict granville s heuristic reasoning which suggests @xmath126{granville}}.\ ] ] * generalized cramr conjecture c. * for any @xmath0 with @xmath12 , we have @xmath127 * generalized cramr conjecture d. * for any @xmath0 with @xmath12 , we have @xmath128 here @xmath123 is the prime at the end of the maximal gap @xmath32 .
the latter generalization of cramr s conjecture seems a little far - fetched .
nevertheless , we are yet to see even a single counterexample with @xmath129 . in february 2016
, conjecture d has been posted at the website _
primepuzzles.net _ @xcite .
this website is frequented by many computation - savvy people ; yet no one came up with a counterexample .
as above , let @xmath130 be the maximal prime gap ending at the prime @xmath123 .
shanks @xcite conjectured that the infinite sequence of maximal prime gaps satisfies the asymptotic equivalence @xmath131{shanks}}.\ ] ] our numerical experiments suggest the following natural generalizations of the shanks conjecture to describe the behavior of @xmath32 .
* generalized shanks conjecture i. * for any @xmath0 with @xmath12 , there exists an infinite sequence @xmath132 that comprises _ almost all _ maximal gaps @xmath32 such that every gap in @xmath132 satisfies the asymptotic equality @xmath133 for @xmath103 and @xmath31 , this formulation is compatible with heuristics of granville @xcite implying that there should exist an _ exceptional infinite subsequence _ of @xmath134 satisfying @xmath135 indeed , our `` almost - all '' formulation simply means that any exceptional subsequence is very thin ; that is , a zero proportion of maximal gaps @xmath134 have exceptionally large sizes predicted by granville .
* generalized shanks conjecture ii .
* _ all _ maximal gaps @xmath32 satisfy @xmath133 this strong formulation contradicts granville s heuristics cited above .
firoozbakht @xcite conjectured that @xmath136 is a decreasing sequence .
equivalently , @xmath137 the conjecture has been verified for all @xmath138 @xcite .
an independent verification was also performed by wolf ( unpublished ) ; see @xcite .
firoozbakht s conjecture implies @xmath139{kourbatov2015u}}.\ ] ] z .- w .
sun ( * ? ? ? * conjecture 2.3 ) generalized firoozbakht s conjecture ( [ firoozbakht ] ) as follows : let @xmath140 be positive integers with @xmath10 odd , @xmath9 even and @xmath141
. denote by @xmath142 the @xmath25-th prime in the progression @xmath10 , @xmath143 , @xmath144 then there exists @xmath145 such that the sequence @xmath146 is strictly decreasing .
in particular , one can take @xmath147 for @xmath148 .
_ if the latter conjecture is true , @xmath145 may be quite large .
for example , with @xmath149 and @xmath150 we must have @xmath151 .
indeed , @xmath152 ( 17010893 and 17163901 are the 1893rd and 1894th primes @xmath153 mod 1168 ) . assuming the grh and reasoning along the lines of @xcite , we can prove that the sun - firoozbakht conjecture implies @xmath154 this in turn implies the generalized cramr conjectures [ ocram ] , [ aacram ] , and [ naivecram ] , as well as a modified form of [ limsupcram ] with @xmath155 . the sun firoozbakht conjecture of sect.[sunfiroozconj ] might be excessively strong .
a weaker , more plausible statement is the `` almost all '' formulation below .
for any coprime @xmath0 , the endpoints @xmath156 , @xmath157 of * _ almost all _ * record gaps @xmath1 satisfy inequality ( [ sunfiroozineq ] ) . here
, again , @xmath142 denotes the @xmath25-th prime in the arithmetic progression @xmath10 , @xmath143 , @xmath144
interested readers can reproduce and extend our results using the programs below .
.... \\ usage example : q=1000;for(r=1,q-1,if(gcd(q , r)==1,maxgap(q , r,1e12 ) ) ) default(realprecision,11 ) \\
li(x ) computes the logarithmic integral of x li(x ) = return(real(-eint1(-log(x ) ) ) ) \\
pmin(q , r ) computes the least prime p = qn + r , for n=0,1,2,3 , ...
pmin(q , r ) = forstep(p = r,1e11,q , if(isprime(p ) , return(p ) ) ) \\ maxgap(q , r , end ) computes maximal gaps g between primes p = qn + r \\ as well as rescaled gap values ( w and h ) .
\\ results are written on screen and in the c:\xgap folder .
\\ computation ends when primes exceed the end parameter .
maxgap(q , r , end ) = { re=0 ; s = pmin(q , r ) ; t = eulerphi(q ) ; sqrtlogphi = sqrt(log(t ) ) ; while(s < end , m = s+re ; p = m+q ; while(!isprime(p ) , p+=q ) ; while(!isprime(m ) , m-=q ) ; g = p - m ; if(g > re , re = g ; lip = li(p ) ; a = t*p / lip ; h = g / a - log(lip / t ) ; w = g / a - log(lip / t)+(0.57*sqrtlogphi+1)*log(p / lip)-2.7*sqrtlogphi+1.2 ; f = ceil(log(p)/log(10 ) ) ; write("c:\\xgap\\1e"f"_"q".txt " , w " " h " " g " " m " " p " q="q " r="r ) ; print(w " " h " " g " " m " " p " q="q " r="r ) ; if(g / t > log(p)^2 , write("c:\\xgap\\1e"f"_"q".txt","extra large " ) ) ; ) ; s = p ; ) } .... .... \\ usage example : q=1000;for(r=1,q-1,if(gcd(q , r)==1,randomgap(q , r,1e10 ) ) ) default(realprecision,11 ) \\
exprv(m ) returns an exponential random variable with mean m exprv(m ) = return(-m*log(random(1.0 ) ) ) \\
randomgap(q , r , end ) writes to c:\ygap\ a set of files with record gaps \\ in a growing sequence of integers p separated by " random " gaps which \\ are exponentially distributed with mean m = phi(q)*log(p ) . \\
the parameter r is included to mimic maxgap(q , r , end ) .
randomgap(q , r , end ) = { re=0 ; p = max(2,r ) ; t = eulerphi(q ) ; while(p < end , g = ceil(exprv(t*log(p ) ) ) ; s = p ; p+=g ; if(g > re , re = g ; lip = real(-eint1(-log(p ) ) ) ; \\li(p ) a = t*p / lip ; h = g / a - log(lip / t ) ; f = ceil(log(p)/log(10 ) ) ; write("c:\\ygap\\rand"f"_"q".txt " , h " " g " " s " " p " q="q " r="r ) ; print(h " " g " " s " " p " q="q " r="r ) ; ) ) } .... to study the distributions of standardized maximal gaps @xmath32 we used the distribution fitting software easyfit @xcite .
data files created with maxgap.gp or randomgap.gp are easily imported into easyfit : from the _ file _ menu , choose _ open _ , select the data file , then specify _ field delimiter = space _ , click _ update _
ok_. * caution : * pari / gp writes real numbers near zero in a mantissa - exponent format _ with a space _ preceding the exponent ( e.g. 1.7874829515 e-5 ) , whereas easyfit expects such numbers _ without a space _
( e.g. 1.7874829515e-5 ) .
therefore , before importing into easyfit , search the data files for `` e- '' and replace all occurrences with `` e- '' .
i am grateful to all contributors and editors of the websites _ oeis.org _ and
_ primepuzzles.net_. thanks also to marek wolf and zhi - wei sun for putting forward many interesting conjectures @xcite .
i am especially grateful to prof.wolf for his very useful comments via email .
r. p. brent , twin primes ( seem to be ) more random than primes , a talk at _ the second number theory down under conference _ ,
newcastle , nsw , australia , oct .
24 - 25 , 2014 .
+ http://maths-people.anu.edu.au/~brent/pd/twin_primes_and_primes.pdf k. ford , b. green , s. konyagin , j. maynard , t. tao , long gaps between primes , arxiv preprint arxiv:1412.5029 ( 2014 ) .
a. granville , harald cramr and the distribution of prime numbers , _ scandinavian actuarial journal _ * 1 * ( 1995 ) , 1228 .
a. kourbatov , is there a limiting distribution of maximal gaps between primes ?
poster presentation at _ dimacs conference on challenges of identifying integer sequences _ , oct .
9 - 10 , 2014 .
a. kourbatov , the distribution of maximal prime gaps in cramr s probabilistic model of primes , _ int .
journal of statistics and probability _ * 3 * ( 2 ) ( 2014 ) , 1829 .
arxiv:1401.6959 a. kourbatov , verification of the firoozbakht conjecture for primes up to four quintillion , _ int .
forum _ * 10 * ( 2015 ) , 283288 . http://arxiv.org/abs/1503.01744 t. oliveira e silva , s. herzog , and s. pardi , empirical verification of the even goldbach conjecture and computation of prime gaps up to @xmath158 , _ math . comp .
_ * 83 * ( 2014 ) , 2033 - 2060 .
sun , conjectures involving arithmetical sequences . in : s.
kanemitsu , h. li , j. liu ( eds . ) _ number theory : arithmetic in shangri - la .
6th china - japan seminar _
( shanghai , august 15 - 17 , 2011 ) , world sci . ,
singapore , 2013 , pp.244258 .
( concerned with sequences http://oeis.org/a005250 [ ] , http://oeis.org/a084162 [ ] , http://oeis.org/a235402 [ ] , http://oeis.org/a268799 [ ] , http://oeis.org/a268925 [ ] , http://oeis.org/a268928 [ ] , http://oeis.org/a268984 [ ] , http://oeis.org/a269234 [ ] , http://oeis.org/a269238 [ ] , http://oeis.org/a269261 [ ] , http://oeis.org/a269420 [ ] , http://oeis.org/a269424 [ ] , http://oeis.org/a269513 [ ] , http://oeis.org/a269519 [ ] . ) | let @xmath0 be coprime positive integers .
we empirically study the record gaps @xmath1 between primes @xmath2 of the form @xmath3 .
extensive computations suggest that @xmath4 almost always ; more precisely , @xmath5 , where @xmath6 is the expected average gap between primes @xmath7 , and @xmath8 is a correction term .
the distribution of @xmath1 near its trend is close to the gumbel extreme value distribution .
however , the question whether there exists a _ limiting distribution _ of @xmath1 is open
. 0.5 cm [ theorem]corollary [ theorem]lemma [ theorem]proposition [ theorem]definition [ theorem]example [ theorem]conjecture [ theorem]remark 2.7 cm * on the distribution of maximal gaps + .1 in between primes in residue classes * 0.7 cm alexei kourbatov + www.javascripter.net/math + [email protected] .2 in |
interstellar dust is fairly cold , emitting in the infrared and submillimeter , and so as major observational facilities have become available at these wavelengths unique data have been gathered and great progress has been made through their analysis .
i describe the basic processes involved in emission by dust and large molecules .
observations of both the emission spectrum and the spatial distribution of the emission are of interest .
i enumerate what data bases are available now and then turn to the `` candy shop , '' wherein one finds an enticing array of powerful new facilities that will become available within the next decade .
the variety of instrumentation offers broad - band photometric information , higher resolution spectroscopy for smaller areas of the sky , and wide - field imaging with increased resolution . thus dust in a range of environments from point - like protostellar environments ,
to nebulae and more diffuse emission can be studied .
the emissivity of a single grain of size @xmath0 and temperature @xmath1 is @xmath2,\ ] ] where @xmath3 is the efficiency factor for absorption or emission and @xmath4 is the planck function .
the combination @xmath5 measures the opacity while @xmath6 determines the spectrum .
appropriate integrations over the size distribution of interstellar dust are needed . for @xmath7 , @xmath8 and so rewriting equation [ eqsingle ] @xmath9.\ ] ]
this is the origin of the statement that infrared or submillimeter emission is a measure of the total mass of dust . of course
, to calibrate this one needs @xmath1 ( usually determined empirically from the broad - band spectrum ) and @xmath10 ( usually from theory , though the spectral dependence is constrained by observations ) .
with multiwavelength imaging one can obtain temperature - compensated maps ( even without a perfect calibration ) of the ( relative ) spatial distribution of column density or opacity .
schlegel , finkbeiner , & davis ( 1998 ) constructed high resolution all - sky maps of optical extinction @xmath11 using iras data ( at 4 resolution ) with lower resolution dirbe and h i 21-cm emission data and colours of cluster ellipticals as steps in the calibration ( see also martin 1994 for simple limited - field application ) .
grain temperature is determined by radiative energy balance between absorption and emission ( see also [ short ] ) . in the milky way
the energy density of starlight and that of the cosmic 3 k microwave background are about the same and so the radiation temperature is about 3.6 k. however , this is directly relevant only for a grain with @xmath12 , a grey or black body .
grains are hotter than this because they absorb starlight ( characteristically optical and ultraviolet ) fairly efficiently ( @xmath13 ) whereas they emit inefficiently ( characteristically infrared ; @xmath14 ) ; the grains warm up , increasing @xmath4 to make up for the low @xmath3 .
the equilibrium @xmath1 is the temperature for which the emission balances the absorption .
a simple illustration is to assume that for all frequencies @xmath15 and that the interstellar radiation field is a blackbody of temperature @xmath16 k diluted by @xmath17 , whence @xmath18 .
inserting @xmath19 gives 3.2 k , as mentioned above .
inserting @xmath20 gives 16 k , very close to that observed in the diffuse interstellar medium .
but it should be pointed out that the latter is a coincidence ; for the sizes of grains that are actually found in the interstellar medium the spectral dependences of emission and absorption give @xmath21 closer to 2 and 0 , respectively , with offsetting effects ( the relative normalization is then important too ) .
obviously , in detail @xmath1 is a function of grain size @xmath0 and composition even for the same radiation field .
some useful examples are given by draine & lee ( 1984 ) .
dust grains close to luminous stars are bathed in a higher than average radiation field , and so are hotter . for example , at a distance of a few pc from an o8 v star , the dust temperature is closer to 30 k. on the other hand , in the quiescent core of a molecular cloud , dust is shielded from starlight .
a factor of four less radiation to be absorbed would drop the temperature by @xmath22 , from say 17 to 14 k. where there is a bright infrared background from warm dust ( or scattered light ) , foreground cooler dust can be imaged in extinction , given sufficient opacity .
because of the spectral dependence of @xmath3 , the opacity is larger at higher frequencies .
a good way to assess where grains emit most of their energy is to look at a plot of @xmath23 ( or equivalently @xmath24 ) . if @xmath25 in the frequency range of emission , then from equation [ eqsingle ] , @xmath26 .
this function peaks at @xmath27 for high latitude cirrus clouds with @xmath28 k , @xmath29 m ( in other units , @xmath30 @xmath31 or @xmath32 hz ) . thus iras , with the longest wavelength band at 100 @xmath33 m ( 83 120 @xmath33 m ) , still missed a substantial fraction of the energy . iras detected widespread diffuse cirrus emission in the shorter wavelength bands , at 60 , 25 , and 12 @xmath33 m and this has been followed up with other experiments ( see fig .
[ fig-1 ] ) .
why should there be emission at @xmath34 ?
the first answer is couched in terms of non - equilibrium emission by very small grains ( vsgs ) .
these have such a small heat capacity that a single ultraviolet photon raises the temperature well above the equilibrium @xmath1 that would be calculated from the straightforward radiative balance described above . to be in this regime
, grains must have @xmath35 , considerably smaller than classical grains ( @xmath36 ) causing optical extinction and polarization , but conceivably part of a continuous size distribution such as deduced from extinction ( kim , martin , & hendry 1994 ) . the grains still emit according to equation [ eqsingle ] , but now @xmath1 is a function of time , initially high but decreasing in a matter of minutes through emission of infrared photons , eventually to below the equilibrium temperature . the temperature distribution has been modeled ( e.g. , guhathakurta & draine 1989 ) .
vsgs are thought to be responsible for the enhanced 60 and 25 @xmath33 m emission ( see model in fig .
[ fig-1 ] ) .
the second answer is more often couched in terms of internal conversion of the absorbed ultraviolet radiation within large molecules , with emission in infrared spectral features ( if considered as tiny grains , then the spectral features come from the spectral dependence of @xmath3 ) .
prominent in the astronomical setting are polycyclic aromatic hydrocarbons ( pahs ) , though the exact species are still poorly characterized .
spectral measurements of widespread diffuse cirrus emission by irts and iso have shown that the 11 @xmath33 m pah feature ( see fig . [ fig-1 ] ) accounts for a large fraction of the energy detected in the iras 12 @xmath33 m band .
note that pahs emit a substantial fraction of @xmath23 , and so they must account for a corresponding fraction of the diffuse starlight absorbed .
pahs are also strong emitters in photodissociation regions ( pdrs ) near hot stars where there is an abundance of ultraviolet radiation ( see fig .
[ fig-2 ] ) .
emission at @xmath37 has been observed with the firas and dirbe experiments on cobe .
early analysis suggested that there was an excess of longer wavelength emission ( @xmath38 @xmath33 m ) , attributed to a relatively cold 7 k dust component ( e.g. , dwek et al .
1997 ) . despite imaginative suggestions ,
there is no convincing explanation why some dust should be this cold .
recent analysis by lagache et al .
( 1998 ) has concluded that the 7 k dust component is spurious , resulting from analyses which neglected an isotropic far infrared background .
this background has a different spectrum and causes significant contamination at long wavelengths ( they also find that the long wavelength galactic dust spectrum is consistent with @xmath39 ) . at even lower frequencies , in the microwave range ( 15 50 ghz )
, there is another excess over the thermal emission expected from dust , but which correlates with 100 @xmath33 m emission .
lazarian ( this conference ; also draine & lazarian 1998 ) attributes this to emission from small spinning dust grains .
this component of emission is another `` foreground , '' possibly polarized , that has to be subtracted in studying the cosmic microwave background ( [ planck ] ) .
there are many sources of infrared and submillimeter data on dust .
much more information that what can be given here is available via the respective web sites , a compilation of which is given below .
to provide points of reference , i have listed here many sites for `` completed '' to `` on - going '' facilities . in the next section
, i touch on a few points for `` future '' missions . 1 .
iras : http://www.ipac.caltech.edu/ipac/iras/iras.html * hires : http://www.ipac.caltech.edu/ipac/iras/hires_over.html
* iga : http://irsa.ipac.caltech.edu/applications/iga/ * miga : http://www.cita.utoronto.ca/@xmath40kerton/ @xmath41 ( see also kerton & martin , this conference ) 2 .
cobe : http://www.gsfc.nasa.gov/astro/cobe/cobe_home.html 3 .
irts : http://koala.astro.isas.ac.jp/irts/irts_e.html
kao : http://ccf.arc.nasa.gov/dx/basket/storiesetc/fsarc004.html and @xmath42 http://jean-luc.arc.nasa.gov/kao/homepage.html 1 .
iso : http://www.iso.vilspa.esa.es/ @xmath43 and @xmath44 http://www.ipac.caltech.edu/iso/iso.html 2 .
msx : http://gibbs1.plh.af.mil/ 1 .
jcmt : http://www.jach.hawaii.edu/jcmt/ * scuba : http://www.jach.hawaii.edu/jcmt/scuba/ 2 .
2mass : http://www.ipac.caltech.edu/2mass/
there are many upcoming missions and facilities of great interest to the study of dust . again , there a web sites to find basic information and keep current . 1 .
wire : http://www.ipac.caltech.edu/wire/
map : http://map.gsfc.nasa.gov/ 3 .
sofia : http://sofia.arc.nasa.gov/ 4 .
sirtf : http://ssc.ipac.caltech.edu/sirtf/ 5 .
iris : http://koala.astro.isas.ac.jp/astro-f/index-e.html 6 .
planck : http://astro.estec.esa.nl/sa-general/projects/planck/ 7 .
first : http://astro.estec.esa.nl/sa-general/projects/first/ iris and also map and planck will survey the whole sky like their predecessors iras and cobe . the others will observe more limited areas with some combination of higher sensitivity , spatial resolution , and spectral resolution .
issues in evaluating and intercomparing the upcoming prospects include size of mirror , field of view , array detectors , imaging and spectroscopic capabilities , sensitivity and noise ( cooling ; backgrounds ) , expected availability , service lifetime , and access to observations and data .
the various web pages provide some useful intercomparisons with predecessors and competitors .
wire , a small 30-cm telescope in nasa s explorer series , is to be launched in the spring of 1999 . because the telescope is cooled ( with solid h@xmath45 )
, the mission lasts only about four months .
the primary goal of wire is to reveal how galaxies evolve with time at infrared wavelengths . during repointing from one galactic pole to another ( about 18% of clock time ) wire
will make sensitive observations toward regions that do not compete with the primary mission .
this time has been awarded to a number of associate investigator ( ai ) programs .
ai data are proprietary for a year and then will be released in an archive .
wire produces images at 12 and 25 @xmath33 m ( see fig .
[ fig-1 ] ) with about 20 resolution over a 33 field .
thus it is possible to map significant portions ( several square degrees ) of interstellar clouds .
a team headed by g. helou with which i am involved will study the cloud structure and relative distribution of emission of the distinct pah and vsg components , providing important information on their evolution and sensitivity to local conditions in a variety of environments ( molecular , translucent , cirrus ) .
dense cores will be detected in extinction .
we shall also examine the evolution of dust in interfaces between h ii regions and molecular clouds ( e.g. , fig . [ fig-2 ] ) .
a follow - on to the kuiper airborne observatory ( kao ) , sofia is a boeing 747-sp aircraft modified to accommodate a 2.5-m telescope . developed by nasa and dlr ,
it is expected to begin flying by mid-2001 and have an operating lifetime of some 20 years with about 160 flights ( about 5 hours actual observing each ) per year .
initially it has four facility class science instruments for the use of the user community ( via peer review ) who will not be required to have extensive knowledge or experience in infrared instrumentation or observing techniques .
hawc ( high resolution airborne wideband camera ) with a @xmath46 array covers the 40 300 @xmath33 m spectral range with three bands centered at 60 , 110 , and 200 @xmath33 m at an image scale of two pixels per airy disk ( fwhm ; diffraction limited resolution about 15 at 110 @xmath33 m ) .
forcast ( a wide - field infrared camera ) will sample at 0.75/pixel over a 3.2 @xmath47 3.2 instantaneous field of view for diffraction limited imaging .
selectable filters provide for continuum imaging in the 16 40 @xmath33 m region .
clearly , the increased spatial resolution of these instruments is coming at the expense of sky coverage .
aires ( airborne infrared echelle spectrometer ) utilizes two - dimensional detector arrays and a large echelle grating to achieve spectral imaging with high angular resolution and sensitivity .
wavelength coverage can be selected within 17 210 @xmath33 m .
flitecam ( first light infrared test experiment camera ) will offer wide - field imaging , high - resolution imaging for observing diffraction - limited images at 3 @xmath33 m , low - resolution grism spectroscopy , and pupil - viewing .
there are also five principal investigator class first light infrared instruments , which are all spectrometers ( those with resolving power @xmath48 are probably less interesting for studies of dust ) .
given the long lifetime of the observatory , it will be possible to make instrumentation improvements taking advantage of new technologies like advanced large format arrays .
sirtf , the last of the `` great observatories '' of nasa , consists of an 85-cm telescope . despite a dramatic reduction in budget , the scientific potential has been maintained .
the mission cryogenic lifetime is 2.5 y , though 5 y is possible .
launch is in late 2001 into a solar earth - trailing orbit .
like hst there will be gto time , but a large portion will be for general observer peer - reviewed proposals .
in addition there is legacy science , a unique category of peer - reviewed programs to be carried out by teams of investigators .
legacy science programs are distinguished as being large ( perhaps more than a thousand hours ) coherent science investigations that produce a scientific data archive which is of general and lasting importance to the broad community ( such data are non - proprietary ) .
these large projects are expected to comprise a significant fraction of sirtf s first year of science operations .
sirtf has three cryogenically - cooled science instruments for imaging and spectroscopy over 3 180 @xmath33 m , with large sensitivity gains by using large - format detector arrays .
mips ( multiband imaging photometer for sirtf ) is comprised of three detector arrays : @xmath49 for imaging at 24 @xmath33 m , @xmath50 at 70 @xmath33 m , and @xmath51 at 160 @xmath33 m ; the fields of view are 5 , 5 , and 0.5 @xmath52 , respectively .
there is a scan mirror to provide mapping of larger fields with very efficient use of telescope time .
the @xmath50 array also takes very low - resolution spectra ( @xmath53 ) from 50 100 @xmath33 m .
irac ( infrared array camera ) has four channels providing simultaneous images at 3.6 , 4.5 , 5.8 , and 8 @xmath33 m ; the @xmath54 arrays give a 5 field of view with 1.2pixels .
irs ( infrared spectrograph ) has two modules providing spectra with @xmath55 over 4 40 @xmath33 m and another two providing @xmath56 over 10 37 @xmath33 m .
the low - resolution modules are long slit designs that allow both spectral and one - dimensional spatial information to be acquired simultaneously on the same detector array .
the high - resolution modules use a cross - dispersed echelle design that gives both spectral and limited spatial measurements on the same detector array .
iris , a 70-cm cooled telescope , is the second infrared astronomy mission ( after irts ) of isas with a launch in early 2003 .
unlike many of the other facilities being described , iris will carry out a full infrared sky survey with fis ( far - infrared surveyor ) .
fis is sensitive in the range from 50
200 @xmath33 m , with angular resolutions of 30 50 and much greater sensitivity than iras .
there are four passbands : 50 70 , 50 110 , 150 200 , 110 200 @xmath33 m ; the longer wavelength coverage is important for diffuse dust ( [ spect ] ) . for pointed observations fis
also incorporates a fourier spectrometer covering the entire range , with @xmath57 at 100 @xmath33 m .
irc ( infrared camera ) has large - format detector arrays to take deep images of selected regions ( field of view 10 ) with spatial resolution about 2 ; three channels observe simultaneously to cover the ranges 1.8 5 , 5 12 , and 10 26 @xmath33 m .
there is also a capability of using grisms for low - resolution ( @xmath55 ) slitless spectroscopy .
i lump these two very distinct esa satellites together because they have been combined for a 2007 `` carrier '' launch and deployment to orbits around the second earth - sun lagrange point ( l2 ) .
planck is a 1.5-m telescope for all - sky mapping of the cosmic microwave background .
full sky coverage is obtained in six months .
interstellar medium studies come as a by - product of having to subtract the galactic foregrounds ( free - free and synchrotron continua as well as dust ) .
wide frequency coverage will aid the separation of the various components which have distinctive spectra .
hfi ( high frequency instrument ) will have spectral coverage with bolometers in six channels ( @xmath58 ) from 350 @xmath33 m to 3 mm ( 860 100 ghz ) at 5 resolution above 200 ghz ( like the issa product from iras ) .
hfi also measures linear polarization in three channels .
the polarized dust foreground in this frequency range will be from thermal emission ( [ long ] ) .
polarized emission from rapidly spinning small grains will be of interest to the lfi , with hemt receivers giving spectral coverage over 3 10 mm ( 100 30 ghz ) in four ( @xmath59 ) channels and spatial resolution as good as 2 at 40 ghz .
note also map ( nasa s microwave anisotropy probe ; much earlier launch in fall 2000 ) which will cover 22 90 ghz with a highest resolution of 13 .
first ( far infrared and submillimeter telescope ) , a 3.5-m passively cooled telescope for pointed studies , has an optimum range 80
670 @xmath33 m for photometry and spectroscopy , considering complementary facilities in space ( sirtf ) and on the ground ( scuba ) .
first is quite sensitive , one rough comparison being a yearly output equivalent to about 1000 sofia flights ( with sofia first light instrumentation ) .
eventually 3/4 of the time will be available to general observer proposals .
there are three planned instruments ( about to be confirmed ) , with cryostat lifetime about 3 y. of most interest to studies of dust are pacs and spire .
pacs ( photoconductor array camera and spectrometer ) simultaneously covers 80 130 @xmath33 m and 130 210 @xmath33 m with @xmath60 arrays providing 1.5 @xmath47 1 and 3@xmath47 2 coverage , respectively , at full beam sampling .
the spectroscopic capability provides @xmath61 .
spire ( spectral and photometric imaging receiver ) has a 4 field for simultaneous photometry ( @xmath62 ) at 250 , 350 , and 500 @xmath33 m using square arrays of bolometers with 32 , 24 , and 16 on a side , respectively .
it includes an fts with a 2 field of view for spectroscopy ( adjustable to @xmath63 1000 at 250 @xmath33 m ) over 200
670 @xmath33 m .
hifi is for high resolution ( @xmath64 @xmath65 ) heterodyne spectroscopy in three bands within 480 2700 ghz ( 625 110 @xmath33 m ) .
ballantyne , d. r. , kerton , c. r. , & martin , p. g. 1999 , , in preparation bernard , j. p. , boulanger , f. , dsert , f. x. , giard , m. , helou , g. and puget j. l. 1994 , a&a , 291 , l5 boulanger , f. , et al . 1996 , , 315 , l325 dsert , f - x .
, boulanger f. , puget j. l. 1990 , a&a , 237 , 215 draine , b. t. , & lazarian , a. 1998 , , 508 , 157 draine , b. t. , & lee , h. m. 1984 , , 285 , 89 dwek , e. , et al .
1997 , , 475 , 565 giard , m. , lamarre , j. m. , pajot , f. , serra , g. 1994 , , 286 , 203 guhathakurta , p. & draine , b. t. 1989 , , 345 , 230 kim , s .- h . ,
martin , p. g. , & hendry , p. d. 1994 , , 422 , 164 lagache , g. , abergel , a. , boulanger , f. , & puget , j .- l .
1998 , , 333 , 709 martin , p. g. 1994 , in infrared cirrus and diffuse interstellar clouds , r. m. cutri and w. b. latter , asp conference series 58 : san francisco , 137 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 | interstellar dust appears in a number of roles in the interstellar medium .
historically , the most familiar one is as a source of extinction in the optical . absorbed optical and ultraviolet light heats the dust , whence infrared ( including near - infrared to submillimeter ) emission , the particular theme of this review . in some distant galaxies , most of the luminosity of the galaxy
is thus converted into the infrared , though in the milky way the situation is not generally as extreme , except in localized regions of star formation .
i briefly review the range of physical conditions in which the dust emits in the interstellar medium and the various sizes of dust probed . of interest
astrophysically are observations of both the spectrum and the spatial distribution of the emission , preferably in combination . in the past
fifteen year probes of dust emission have advanced significantly : through iras , kao , cobe experiments , irts , iso , and msx .
satellite and stratospheric observations of dust emission are complemented by ground - based studies , such as imaging and polarimetry in the submillimeter with the jcmt or imaging of reflected light from dust in the near infrared with 2mass .
looking ahead , the next decade promises to be equally exciting .
i give an overview of some of the distinctive features of facilities anticipated in the near term ( wire ) , early in the new millennium ( sofia , sirtf ) , and somewhat further out ( iris , first , planck ) . |
transport properties of small conducting structures are strongly influenced by size effects .
oscillation of magnetoresistance in thin metallic films , and quantization of conductance in narrow wires and point contacts are examples of such effects .
size effects in superconducting tunneling have attracted attention since early experiments by tomasch @xcite . in these experiments , oscillations of the tunnel conductance as a function of applied voltage were found for tunneling from a superconductor to a thin superconducting film of an ns proximity bilayer .
the geometric resonance nature of the effect was clearly indicated by the dependence of the period of oscillations on the thickness of the superconducting film .
similar conductance oscillations for tunneling into a normal metal film of ns bilayers were reported by rowell and mcmillan @xcite . later on an even more pronounced effect
steps on the current - voltage characteristics of sins junctions at applied subgap voltages , @xmath5 was observed by rowell@xcite ( for a review see ref . ) .
in addition to the dependence on the thickness of the n - film , the period of the current steps also shows temperature dependence which scales with the temperature dependence of the superconducting gap @xmath3 .
the current steps occur at applied subgap voltages , @xmath5 , and they are understood as resonant features due to quasiparticle tunneling through superconducting bound states existing in ins wells at energies lying within the superconducting gap , @xmath6 , de gennes saint - james levels@xcite .
recently , properties of superconducting bound states have attracted new attention in connection with observations of conductance anomalies in mesoscopic ns structures .
resonant oscillations of the subgap conductance in mesoscopic quasi - ballistic ns junctions have been reported by morpurgo et al@xcite .
the zero bias conductance peak , a related resonant phenomenon in diffusive ns junctions , was discovered even earlier@xcite and then intensively studied both theoretically and experimentally ( for a review see ref . ) . in these recent studies of mesoscopic junctions ,
attention was switched from the single - particle current through superconducting bound states to the two - particle ( andreev ) current .
the traditional view of subgap current transport in proximity nins and sins structures considers single - particle tunneling into bound states in the normal region of the ins well@xcite , implicitly assuming that the normal region of the ins well plays the role of equilibrium reservoir .
such a model is appropriate for low - transmission tunnel junctions with low tunneling rate compared to the inelastic relaxation rate .
however , transparent mesoscopic structures are in a different transport regime where the bound levels are well decoupled from the superconducting reservoirs , and where injected quasiparticles escape from the ins well via andreev reflection.@xcite resonant two - particle current in quantum nins junctions has been theoretically studied in refs .. in sns junctions , the quasiparticles may undergo multiple andreev reflections ( mar ) before they escape into the reservoirs.@xcite in a number of recent experiments with ballistic sns devices fabricated with high mobility 2d electron gas ( 2deg ) @xcite the highly coherent mar transport regime has been realized .
the purpose of this paper is to develop a theory of coherent multiple andreev reflections which will be applicable to 2deg - based sns junctions . in 2deg devices
the separation of the superconducting electrodes @xmath7 is typically larger than 200 nm , which is of the same order of magnitude as the superconducting coherence length , @xmath8 ( @xmath9 is the fermi velocity of the 2d electrons ) , and superconducting bound states are formed well inside the energy gap .
the presence of bound states in the junctions of finite length gives rise to resonances in the mar transport , which dramatically affects the subgap current .
furthermore , it is possible in 2deg devices to use electrostatic gates to reach the quantum transport regime with a small number of electron modes and variable transmissivity .
a theory of coherent mar has earlier been developed for short superconducting junctions,@xcite @xmath10 , where superconducting bound states do not play any significant role.@xcite such a theory is consistent with the physical situation in atomic - size superconducting point contacts , @xcite where quantization of conduction modes has turned out to be very helpful for detailed comparison between theory and experiment .
the purpose of our study is to investigate the interplay between superconducting bound state resonances and coherent mar in long sns junctions , @xmath11 , in the quantum transport regime . in a number of publications ,
the coherent mar approach has been applied to long sns junctions@xcite . however
, these studies were restricted to fully transparent junctions where the bound states are strongly washed out and the resonances are not pronounced ( in fact , as we will show , at zero temperature the current in such junctions does not show any structures ) .
we will study junctions with arbitrary transmissivity , @xmath12 and pay special attention to the low - transparency limit , @xmath13 , where the resonance effects are most pronounced .
the paper is organized as follows . in sec .
[ sninsjunc ] we derive a 1d model for a gated ballistic 2deg device with one transport mode . in sec .
[ details ] we construct a scheme for calculating mar amplitudes in terms of wave propagation in energy space . in sec .
[ lowtransp ] , single current resonances are studied , and sec .
[ interres ] is devoted to a discussion of the interplay between resonances in multi - particle currents .
the properties of the total subgap current is discussed in sec .
[ discussion ] .
we consider an sns junction similar to the one discussed by takayanagi et al.@xcite schematically shown in fig .
[ s2degs - dev ] .
the junction consists of a normal conducting channel fabricated with a high mobility 2deg , which is confined between superconducting electrodes .
the distance between the electrodes is comparable to the superconducting coherence length and small compared to the elastic and inelastic mean free paths and to the normal electron dephasing length .
the superconductor-2deg interfaces are highly transmissive , the transmission coefficient typically exceeding a value 0.75 , and the number of conducting modes in the 2deg channel is controlled by a split gate .
under these conditions , electrons ballistically move from one electrode to the other while occasionally being scattered by rare impurities or junction interfaces . under a voltage bias applied to the junction , the transport regime corresponds to fully coherent multiple andreev reflections ( mar ) . to calculate the dc current
we will apply the scattering theory approach @xcite generalized for superconducting junctions ; see refs . and references therein . the normal electron propagation through the junction is generally described by the @xmath14-channel scattering matrix . by assuming the split gate to select only one transport mode
, we will characterize the transport through this mode by energy - dependent transmission amplitude , @xmath15 , and reflection amplitudes , @xmath16 ( the energy @xmath17 is counted from the fermi energy ) .
the scattering amplitudes satisfying the unitarity relations , @xmath18 , @xmath19 .
the energy dispersion of the scattering amplitudes will introduce the normal electron ( breit - wigner ) and superconducting ( andreev ) resonances in the scattering problem .
the effect of narrow breit - wigner resonances on coherent mar was earlier studied by johansson et al.@xcite and levi yeyati et al.@xcite here we will focus on the effect of andreev resonances and only consider breit - wigner resonances which are wide on the scale of the energy gap .
this will allow us to neglect the energy dispersion of the junction transparency , @xmath20const . however , the scattering phases may depend on the energy , which yields the andreev resonances . in the simplest case , this dependence is a linear function within the energy interval @xmath21 , and we will write it on the form @xmath22 where @xmath23 are constant . in this case , the scattering properties of the normal channel are similar to those of a 1d nin junction . indeed , the corresponding 1d transfer matrix , @xmath24 can be decomposed into a product of three transfer matrices , @xmath25 where @xmath26 is a pauli matrix .
the first and the last matrices describe ballistic propagation of an electron , with wave vector @xmath27 , through the right and left n - regions of an effective junction with lengths @xmath28 and @xmath29 respectively ( from right to left ) , and the matrix @xmath30 describes an effective barrier ( i ) .
quasiparticle propagation through the effective 1d snins junction is described by means of the time - dependent bogoliubov - de gennes ( bdg ) equation , @xcite @xmath31 where @xmath32 is the normal electron hamiltonian , @xmath33 is the impurity potential and @xmath34 is the applied voltage .
the superconducting order parameter @xmath35 is constant within the superconducting electrodes and zero within the normal region @xmath36(see fig .
[ potentialdiag ] ) . in the further calculations ,
the impurity potential is described by the transfer matrix @xmath37 in eq .
( [ t ] ) .
the spatial distribution of the applied potential along the channel is modeled with a step - like function @xmath38 .
in fact , the actual spatial distribution of the potential does not play any role in this system : it can be included in the transfer matrix in eq .
( [ t ] ) , leading to an additional energy - independent shift in the scattering phases in the matrix @xmath37 . as we will see later [ comment after eq .
( [ resphase ] ) ] , the energy - independent phases in the @xmath37-matrix do not affect the current , and can therefore be excluded .
the phase difference between the two superconductors follows from the josephson relation ( @xmath39 ) and introduces time dependence into the problem .
the superconducting electrodes are considered to be equilibrium reservoirs where the quasiparticle wave function is a superposition of electron- and hole - like plane waves , @xmath40 in this equation , the @xmath41 signs in the time - dependent factors refer to the left / right electrode , @xmath42 is the wave vector of electron / hole - like quasiparticles , and @xmath43 , @xmath44 are the bogoliubov amplitudes .
the ratio of the bogoliubov amplitudes equals the amplitude of andreev reflection for particles incoming from the neighboring normal region , @xmath45 since the time dependencies of the wave functions in the two reservoirs are different , the quasiparticle scattering by the junction is inelastic , and one has to consider a superposition of plane waves with different energies in order to construct scattering states .
we will now proceed with the construction of recurrences for the scattering amplitudes following the method suggested by johansson et al .
@xcite . to this end
we introduce the wave functions in the left / right normal region ( @xmath46 ) of the junction with respect to the position of the impurity .
a particular scattering state , labeled with the energy @xmath17 of the incoming quasiparticle , will consist of a superposition of plane waves with energies @xmath47 , where @xmath48 is an integer , @xmath49 , @xmath50\ ] ] @xmath51.\ ] ] the normal electron / hole wave vector @xmath52 is here defined as @xmath53 , @xmath54 .
the meaning of the labels for the scattering ( mar ) amplitudes @xmath55 will be explained below .
continuity of the scattering state wave function across the left and right ns interfaces determines the relation between the electron and hole amplitudes in the vicinity of each interface , @xmath56 which describes elastic andreev reflection (
l / r indices are omitted ) .
it is convenient to consider scattering amplitudes near the impurity ( at @xmath57 ) rather than at the ns interfaces , and to rewrite eq .
( [ andrreflection ] ) for such amplitudes , combining the amplitudes of the ballistic propagation through the normal regions with the andreev reflection amplitude . then , in vector notation , @xmath58 the modified relation ( [ andrreflection ] ) takes the form @xmath59 where @xmath60 the phase @xmath61 , characterizing @xmath62 , is real inside the energy gap , @xmath63 , where it describes the total energy dependent phase shift due to ballistic propagation and andreev reflection .
outside the gap , @xmath64 , the phase @xmath65 has an imaginary part which describes leakage into the superconducting reservoirs due to incomplete andreev reflections . by matching harmonics with the same time dependence in eq .
( [ ansatz ] ) , we derive a relation between scattering amplitudes at the left and the right side of the barrier : @xmath66 where the effective barrier transfer matrix @xmath67 is defined in eqs .
( [ te1 ] ) , ( [ t ] ) . the recursion relations in eqs .
( [ urelation ] ) and ( [ trelation ] ) couple the scattering amplitudes @xmath68 into an infinitely large equation system .
this equation system describing coherent mar is illustrated by the mar diagram in fig .
[ coeffpic ] .
the electron part of the quasiparticle injected at the left ns interface propagates upwards along the energy axis , the amplitudes for this propagation being labeled with @xmath69 . at the injection energy @xmath70 ( amplitude @xmath71 ) , the quasiparticle is accelerated across the barrier ( i ) , where the potential drops .
thus , it enters the right normal part of the junction with energy @xmath72 ( @xmath73 ) , undergoes andreev reflection and goes back as a hole ( @xmath74 ) , entering the left normal part of the junction having been accelerated to energy @xmath75 ( @xmath76 ) , and is then again converted into an electron ( @xmath77 ) .
the @xmath41 indices label the amplitudes after ( @xmath78 ) and before ( @xmath79 ) the andreev reflection for propagation upwards along the energy axis .
there is a similar trajectory of injected holes , which descends in energy , with the mar amplitudes labeled with @xmath80 .
due to electron back scattering at the barrier , the upward and downward propagating waves are mixed , e.g. @xmath71 being not only forward scattered into @xmath73 , but also back - scattered into @xmath81 , which opens up the possibility of interference .
injection from the left reservoir , shown in fig .
[ coeffpic ] , generates a mar path which only connects even side bands at the left side of the junction with odd side bands at the right side .
injection from the right reservoir will generate a different mar path , with even side bands at the right side of the junction , i.e. the diagram in fig .
[ coeffpic ] will effectively be mirrored around the barrier ( i ) .
thus , there are two independent equation systems for the mar amplitudes : injection from the left and from right .
the @xmath82 labels in the mar amplitudes can then be omitted since they are uniquely defined by the source term and the side band index .
the transport along the energy axis generated by mar , from energy @xmath17 to @xmath83 , is conveniently described by the effective transfer matrix @xmath84 , @xmath85^{-1}\hat { c}_{0- } , \ \
n<0\ ] ] @xmath86 where @xmath87 and @xmath88 for the injection from the left ( for the injection from the right , the even and odd side band indices are interchanged ) . for paths within the superconducting gap , @xmath89
, the matrix @xmath84 satisfies the standard transfer matrix equation , @xmath90 , which provides conservation of probability current along the energy axis , @xmath91 an important consequence of the coherence of mar is the possibility of transmission resonances in energy space . from the form of the @xmath92-matrix ,
@xmath93 it is evident that when @xmath94 , the two matrices @xmath95 and @xmath67 will cancel each other , and the probability of transmission through this part will be unity , which leads to resonant enhancement of mar .
the solutions @xmath96 of the resonance equation , @xmath97 coincide with the spectrum of the de gennes saint - james levels localized in ins quantum wells @xcite . the corresponding bound states are located either on the left or the right side of the junction . without loss of generality ,
the calculations can be performed with a real matrix @xmath67 .
the transformation to such a real matrix is given by @xmath98 , with diagonal unitary matrices @xmath99 whose elements are constructed with the scattering phases , which are energy - independent .
it is clear from eq .
( [ mresstruc ] ) that since these energy - independent matrices commute with the matrices @xmath100 , they cancel each other , and the matrix @xmath84 undergoes a similar transformation .
this will lead to an overall phase shift of the scattering state which does not affect the current .
it is interesting to consider the special case of fully transparent junctions , @xmath101 , which has been studied in the literature @xcite . in this case , all matrices @xmath102 in eq .
( [ mdef ] ) are equal to the unity matrix , and the @xmath92-matrix takes the simple form @xmath103 .
the length of the junction then enters only through the phase of the mar amplitudes , which drops out of the side band current .
thus the dc current of fully transparent sns junctions is independent of length and equal to the current in quantum constrictions @xcite . in particular , at zero temperature this current
does not show any structures in the subgap current .
it is also worth mentioning that in this particular case of fully transparent sns junctions , the @xmath92-matrix is diagonal and therefore a closed set of recursive relations can be derived for the mar probabilities ( not just for the mar amplitudes , as in the general case ) , equivalent to the equations for distribution functions derived in the original paper by klapwijk , blonder and tinkham @xcite .
equation ( [ mar ] ) describes `` source - free '' propagation along the mar ladder . to complete the set of equations for the mar
amplitudes we need to take into account quasi - particle injection , which introduces a source term in eq .
( [ mar ] ) . to this end , let us consider a quasiparticle incoming from the left superconducting electrode with energy @xmath17 , having a wave function of the form @xmath104 .\ ] ] the two terms in this equation refer to electron- ( @xmath105 ) and hole - like ( @xmath106 ) injected quasiparticles .
we now include this wave function into the continuity condition at the ns interface at energy @xmath17 , which gives us the following relation between the mar amplitudes @xmath107 and @xmath108 , @xmath109 @xmath110 for quasiparticles injected from the right , a similar equation holds with the substitutions @xmath111 , and @xmath112 .
equations ( [ mar ] ) and ( [ source ] ) give a complete set of equations for the mar amplitudes with the boundary conditions @xmath113 at infinity .
a formal solution of eqs .
( [ mar ] ) and ( [ source ] ) , which is useful both for numerical calculations and analytical investigations , can be constructed by reducing this infinite set of recursion relations to a finite set by representing the mar process above @xmath114 and below @xmath115 by boundary conditions involving reflection amplitudes @xmath116 and @xmath117 , defined as @xmath118 and @xmath119 .
this gives the following representation for the vectors in eq .
( [ coeffdef ] ) , @xmath120 the reflection amplitudes @xmath116 and @xmath117 are independent of the injection , in contrast to the coefficients @xmath121 .
furthermore , they are determined by the boundary conditions @xmath113 and can be expressed in terms of the matrix elements of @xmath122 and @xmath123 , where @xmath124 , @xmath125^{-1}\left ( \begin{array}{c } r_{0- } \\ 1 \end{array } \right)=0.\ ] ] in other words , the vectors in eq .
( [ lim ] ) are equal to the asymptotical values of the eigenvectors of @xmath126-matrices corresponding to the eigenvalues which decrease when @xmath14 goes to infinity .
the advantage of introducing the reflection amplitudes @xmath116 and @xmath117 is that although they have to be calculated numerically , the recurrences which they obey do not contain resonances and converge rather quickly .
this is in contrast to the matrix @xmath84 , which does possess resonances , but which can be calculated analytically in a straight forward way for any given @xmath48 .
the solutions of the recursion equations eqs .
( [ urelation ] ) and ( [ trelation ] ) can now be explicitly written down .
for any given energy @xmath17 we get four different sets of solutions for four scattering states including electron / hole injection from the left and the right . using the formal expression in eq .
( [ nullvectors ] ) and the matrix elements of @xmath127 , the solutions for injection from the left ( @xmath128 ) have the form , @xmath129 } { m_{22}+m_{21}r_{0-}e^{2i\varphi_{0 } } -m_{12}r_{n+}e^{2i\varphi_{n } } -m_{11}r_{0-}r_{n+}e^{2i\varphi_{0}+2i\varphi_{n } } } \left(\begin{array}{c } 1\\ r_{n+ } \end{array}\right).\ ] ] the solutions for injection from the right can be found through interchanging @xmath130 and calculating all quantities with respect to injection from the right .
the solutions for @xmath131 are calculated in a similar manner . now turning our attention to the current , we calculate it in the normal region next to the barrier , using the wave function in this region , @xmath132 , and assuming quasiparticle equilibrium within the electrodes .
the current then takes the form , @xmath133 where @xmath134 is the superconducting density of states , and the sum is over the four scattering states at a given energy @xmath17 associated with the electron- and hole - like quasiparticles ( e / h ) injected from the left and right ( l / r ) .
the current can be divided into parts with different time dependence and expressed as a sum over harmonics : @xmath135 focusing on the dc ( @xmath136 ) component , and calculating the contribution of each scattering state at the injection side of the junction , we express the current spectral density @xmath137 through the probability currents of electrons and holes at energies @xmath138 ( fig .
[ coeffpic ] ) , @xmath139 @xmath140 these currents coinside with the probability currents @xmath141 , eq.([jp ] ) , flowing along the energy axis .
it is convenient to introduce a leakage current @xmath142 , defined as the difference of the probability currents before and after andreev reflection , @xmath143 @xmath142 represents the amount of probability current from all the scattering states injected at energy @xmath17 and leaking out of the junction at energy @xmath114 ( fig . [ coeffpic ] ) . the leakage current is zero inside the energy gap due to complete andreev reflection , @xmath144 , @xmath145 [ cf .
( [ jp ] ] . the explicit expression for the leakage current for @xmath146 follows from eq . ( [ leakage ] ) after insertion of eqs . ( [ coeffsol ] ) and ( [ urelation ] ) , @xmath147 it follows from eq . ( [ ipn ] ) that the leakage currents are positive for all @xmath146 , @xmath148 .
one can also show that they satisfy the inequality @xmath149 , which is a consequence of the conservation of probability current : the leakage current of all side bands except of the side band @xmath150 does not exceed the probability current injected into four scattering states .
furthermore , the leakage current satisfies the important detailed balance equation @xcite , @xmath151 i.e. the leakage at energy @xmath152 due to the injection at energy @xmath17 is the same as the leakage at energy @xmath17 due to injection at energy @xmath152 . using the continuity of current across the barrier , @xmath153 , guaranteed by the transfer matrix @xmath67
, we can express the probability currents in eq .
( [ dccurrent ] ) through the leakage current , eq .
( [ leakage ] ) , @xmath154 @xmath155 by adding and subtracting consecutive terms in the sum . the spectral density of the dc charge current eq .
( [ dccurrent ] ) can then be written on the form @xmath156 since @xmath142 appears in @xmath48 probability currents .
this formula has a clear physical meaning : the contribution to the charge current of the @xmath48-th side band is proportional to the leakage current of the side band times the effective transferred charge @xmath157 .
the detailed balance of the leakage currents , eq .
( [ cancel ] ) , allows us explicitly to prove that at zero temperature the scattering processes between ( occupied ) states with negative energies , @xmath158 do not contribute to the current , in agreement with the pauli exclusion principle .
indeed , by separating the contributions from side bands with @xmath131 and remembering that the leakage current is zero within the gap , we get for zero temperature , @xmath159 where the first and the third terms cancel each other by virtue of eq .
( [ cancel ] ) . at finite temperature , these two terms produce current of thermal excitations while the second term gives the current of real excitations created by the voltage source . keeping only this term which dominates at low temperature
, we finally get , @xmath160 we end this section by noting a technically useful symmetry in the current density , namely @xmath161 , seen from the explicit form of the @xmath84-matrix .
this allows us to reduce the integration interval in eq .
( [ npartcurrent ] ) to @xmath162 .
the approach formulated above provides necessary foundations for numerical calculation of the current for arbitrary transparency and length . however , to get a full understanding of the rich subgap structure in the current - voltage characteristics , which may seem quite random , especially for intermediate transparencies and lengths ( see figs .
[ itotd1plots]-[itotl4plots ] ) , we will conduct a detailed analytical study of the limit of low transparency @xmath13 .
the separation of currents into @xmath48-particle currents , eq . ( [ npartcurrent ] ) , is our basis for analysis , and we will study each current @xmath163 separately . as explained in the previous section , the de gennes saint - james levels , eqs .
( [ mresstruc ] ) and ( [ resphase ] ) , are important for the current transport through the junction , leading to resonant enhancement of the current .
our main attention in this and the next section is on the calculation of the position , height and width of the main current peaks and oscillations which have the magnitude of order @xmath2 . to simplify notations ,
left / right injection indices are omitted in most cases .
the single - particle current , which dominates at large applied voltages , has , according to eq .
( [ npartcurrent ] ) , an onset at @xmath164 .
the full numerical solution for the single - particle current is plotted in fig .
[ i1d1plots ] .
the current shows pronounced oscillations , and the magnitude of the slope at the current onset strongly depends on the junction length . to understand this behavior , we analyze eq .
( [ ipn ] ) in the limit of small transparency @xmath0 , i.e. in the tunnel limit .
first we note ( see appendix [ refexp ] ) that the reflection amplitudes @xmath116 and @xmath117 may be expanded as @xmath165 @xmath166 after inserting the explicit form of @xmath167 together with the expansion ( [ reflectionexp ] ) into eq .
( [ ipn ] ) and putting @xmath168 , we can write the single particle current on the form @xmath169 \tanh(|e|/2k_bt),\ ] ] where @xmath170 in analogy with the tunnel formula for the current , @xcite @xmath171 is identified as the tunneling density of states ( dos ) on the left / right side of the junction . in fig
. [ densplot ] the energy dependence of the dos is presented .
the deviation of this dos from the normal metal density of states is a manifestation of the proximity effect .
the expression ( [ densityofstates ] ) for the dos has earlier been derived for proximity ns sandwiches @xcite .
note that the dos in our case is constant throughout the n - regions . in junctions with arbitrary length ,
the dos usually approaches zero at the gap edge @xmath172 ( ref . ) .
exceptions are junctions with lengths @xmath173 where a bound state splits off from the gap edge . in this case , the dos diverges at the gap edge . the quantum well structure of the sns junctions also give rise to quasi - bound states in the continuum spectrum , @xmath174 , seen as oscillations in the dos .
the single particle current in eq .
( [ i1smalld ] ) is written as the integral over the product of the dos at the entrance energy @xmath17 and the exit energy @xmath175 .
the latter depends on the applied voltage , as well as the integration interval , and therefore the dos oscillations produce oscillations of the current @xmath176 as a function of voltage ( rowell - mcmillan oscillations @xcite ) .
the oscillations become more pronounced when the junction is sufficiently long , and the differential conductance may even become negative .
it is also clear that the dos oscillates as a function of the length of the junction , which give rise to oscillations also in @xmath176 . in short junctions , @xmath177 ,
the current onset at @xmath164 is very steep , see fig .
[ i1d1plots ] . in junctions with finite length
, the current onset is smeared and replaced with a smooth oscillating behavior .
this can be directly related to the smearing of the singularity in the dos at the gap edge .
the length where the crossover between these two behaviors occurs can be taken as a measure of when finite length effects become important . to estimate this length ,
we write equation ( [ ipn ] ) for small lengths @xmath178 , near the threshold , @xmath179 , @xmath180 , keeping the first order terms in @xmath2 in the denominator . for a symmetric junction , @xmath181
, we get @xmath182 from this formula it is clear that for short junctions ( @xmath183 ) , the current onset has the width @xmath184
. if @xmath7 is of the order of @xmath185 , the size of the onset has substantially diminished , and there is no visible onset at @xmath164 when @xmath186 .
this crossover between steep onset and smooth behavior , which happens already for quite short lengths if @xmath2 is small , can be interpreted in terms of a bound state which is situated exactly at the gap edge in short junctions ( @xmath183 ) , and which moves down into the gap when @xmath187 , the effect becoming fully pronounced when the distance from the gap edge , @xmath188 , exceeds the dispersion of the andreev state , @xmath189 , in symmetric junctions@xcite . when @xmath190 approaches @xmath191 , the lowest quasi - bound state in the continuum spectrum approaches the gap edge .
this leads to an accumulation of the spectral weight at the gap edge and reappearance of the singularity in the dos , which results in the reappearance of a sharp current onset at @xmath164 , but with smaller magnitude ; see fig .
[ i1d1plots ] ( @xmath192 ) .
it is of interest to note that in our calculations , based on the scattering theory approach , the bound states are not directly involved in the single - particle transport , which therefore is non - resonant and shows no subgap resonance peaks . within the tunnel model
approach the situation is qualitatively different : the dos in eq .
( [ i1smalld ] ) usually includes the contribution of the broadened bound states , and therefore the single - particle current exists and has pronounced resonant features at subgap voltages @xmath5 .
this difference results from the fact that , within the tunnel model approach , the superconducting bound states are implicitly assumed to be connected to the reservoirs ( broadening due to inelastic interaction ) , which allows a stationary current to flow through the bound states .
in contrast , within the scattering approach , the bound states are disconnected from the reservoirs and have zero intrinsic width . in this case
the bound states obtain their width only due to higher order tunneling processes involving andreev reflections , which are manifested by the resonant multi - particle currents . in practice ,
the relevance of the multi - particle versus single - particle mechanism of the subgap current transport is determined by physics and depends on the ratio of the corresponding dwelling and relaxation times@xcite . in this paper ,
the inelastic relaxation time @xmath193 which determines the width of the single - particle resonances is assumed to be much larger than the dwelling time of the most important two - particle current , @xmath194 .
the two - particle current @xmath195 in quantum point contacts , ( @xmath178 ) is of order @xmath196 when @xmath5 and of order @xmath197 when @xmath198 ( ref . ) .
for finite length junctions , the situation is different .
for the mar paths where the energy of the andreev reflection coincides with a bound state , the current spectral density @xmath199 is of order unity , due to resonant transmission through this state .
for low transparency @xmath0 , this gives a sharp concentration of the current density around the resonant energies . in this limit ,
the two - particle current is well described by the sum of contributions from these resonances , and to evaluate them we examine the energy dependence of @xmath200 close to the resonant energies , @xmath201 .
let us consider the contribution to the leakage current @xmath202^{l}$ ] from quasi - particles injected from the left . as shown in appendix [ twopartres ] , in this case eq . ( [ ipn ] ) reduces to the standard breit - wigner resonance form @xmath203^{l}=\frac{\gamma_{0}^{(m ) } \gamma_{2}^{(m)}}{\displaystyle \left(\frac{\delta e-\delta e^{(m)}}{\delta}\right)^{2}+ \left(\frac{\gamma_{0}^{(m)}+\gamma_{2}^{(m)}}{2 } \right)^{2 } } , \ ] ] where the tunneling rates @xmath204 are given by @xmath205 , @xmath206 , and @xmath207 and the position of the resonance is shifted by @xmath208 an analogous result is valid for quasiparticles injected from the right .
after integrating over energy , the two - particle current in the resonance approximation may be written on the form @xmath209 where the summation is over the positive bound level energies , @xmath210 , and the dos @xmath211 should be calculated at the injection side of the junction and @xmath212 $ ] . according to eq .
( [ i2approx ] ) , the two - particle current @xmath213 increases in a step - like manner in the voltage region @xmath214 .
the steps occur at every voltage where a new resonant channel through a bound state opens up , at @xmath215 .
we note that the step positions depend on temperature and approximately scale with @xmath3 .
each current step has the height of order @xmath2 . as seen from eq .
( [ i2approx ] ) , the contribution to the current of a particular bound state @xmath96 is modulated , as a function of voltage , by the oscillations of the density of states at the entrance and exit energies , @xmath216 . in other words , the pronounced oscillations of the two - particle current seen in fig .
[ i2d1plots ] reflect how close the entrance and exit energies @xmath217 are to a quasi - bound state in the continuum . for @xmath198 ,
the two - particle current @xmath195 oscillates around a constant value with an amplitude of oscillation decreasing as @xmath218 for large voltages .
it is interesting to compare the resonant structures of the two - particle current with the resonant structures in nins junctions.@xcite in nins junctions , the resonant current steps occur at @xmath219 , and they do not have any modulation because the dos on the normal side of the junction is constant . the distance between the resonances and the resonance widths are proportional to the bound level spacing , and they decrease in long junctions . for sufficiently long junctions
, the two - particle current may thus give the appearance to include a series of peaks , as shown on fig .
[ i2comp ] .
in symmetric junctions , the bound state energies at both sides of the barrier will coincide , reducing the number of steps with a factor of two , and giving current steps of double height .
we will conclude this subsection by noting that the difference between the full numerical calculation of the two - particle current and the resonant approximation given in eq .
( [ i2approx ] ) is rather small already when @xmath220 , as can be seen in fig .
[ i2comp ] .
excess current in sns junctions , i.e. the difference between the current in superconducting junction and in the normal junction at large voltage , @xmath221 is commonly considered as a measure of the intensity of andreev reflection . in tunnel sis junctions and low - transmissive point
contacts the excess current is small , @xmath222 , @xmath13 , while in fully transparent contacts the excess current is large , @xmath223 , @xmath101.@xcite accordingly , one would expect large excess current in long sns junctions due to the resonant enhancement of the two - particle current .
however , the excess current is small because of a large deficiency , of order @xmath2 , of the single - particle current caused by the broadening of the current onset at the threshold .
as we will show , the single - particle and two - particle currents undergo a fine cancellation , yielding small net excess current of order @xmath196 when @xmath13 . the excess current has contributions only from the single- and two - particle currents , since all higher order currents include at least one andreev reflection outside the gap whose probability is of order @xmath224 . in the limit of large voltage , @xmath225 , the relevant part of the current in eq .
( [ npartcurrent ] ) then takes the form @xmath226 @xmath227 these equations are written for symmetric junctions , @xmath228 , and for zero temperature ; small andreev reflection amplitudes @xmath229 have been neglected in eq .
( [ ipn ] ) .
the behavior of the current in eq .
( [ ihighvoltage ] ) as a function of voltage is presented in fig .
[ excessapproach ] for different lengths .
it is clearly seen that the limiting value of the excess current is approached much faster in finite length sns junctions compared to point contacts ( @xmath183 ) . in fig .
[ excessplot ] the excess current behavior with respect to the junction length is presented for different transparencies . to analytically examine the excess current in the limit of small transparency , @xmath13 , it is convenient to start with equations ( [ i1smalld ] ) and ( [ i2approx ] ) . to first order of @xmath2
the excess current assumes the form ( @xmath230 ) , @xmath231 @xmath232de,\ ] ] @xmath233 let us consider the contributions to the single - particle current from the left electrode , @xmath234^{l}=-\frac{2ed\delta}{h } + \frac{2ed}{h}\int_{\delta}^{\infty}\left [ n^{l}(e)-1\right]de.\ ] ] inserting @xmath235 from eq .
( [ densityofstates ] ) , this equation can be transformed to the form , @xmath236^{l}= \frac{2ed}{h}\int_{\delta}^{\infty}\left(\frac{e\xi } { \xi^{2}+\delta^{2}\sin^{2}(2el_{l}/\delta\xi_{0})}- \frac{e}{\xi}\right)de= -\frac{ed}{h}\int_{-\infty}^{\infty}d\xi \frac{\sin^{2}(2el_{l}/\delta\xi_{0})}{\xi^{2}+\sin^{2 } ( 2el_{l}/\delta\xi_{0})},\ ] ] where @xmath237 .
it is now possible to analytically continue the integral in the upper half plane which will reduce the integral to a sum over the residues of the poles given by the equation @xmath238 . comparing this equation with eq .
( [ resphase ] ) we find that the poles coincide with the energies of the bound states in the gap .
the excess current contribution from the left - injected single particle current is thus @xmath236^{l } = -\frac{2d\pi e\delta}{h}\sum_{m\geq0 } \frac{1}{\eta^{(m)}}=-\left[i^{exc}_{2}\right]^{l},\ ] ] where @xmath239^{l}$ ] is the contribution to the two - particle current from the bound state resonances at the left electrode .
a similar relation is derived for current from the right electrode .
thus , there is exact cancellation of the excess single - particle and two - particle currents to first order in @xmath2 .
it is interesting to note that the cancellation effect is related to the conservation of the number of states in a proximity normal metal compared to the conventional normal metal .
it follows from eq .
( [ i1excl ] ) that @xmath240 is equal to the difference between the number of continuum states in the proximity metal and the total number of states in a conventional metal , while , on the other hand , the number of the bound states , is equal to @xmath241 according to eq .
( [ iexcsmalld ] ) .
for processes with several andreev reflections ( @xmath242 ) , the possibilities for resonances increase
. every andreev reflection energy may coincide with a bound state energy and thus be resonant .
for some specific voltages , more than one resonance is important , creating a situation of overlapping resonances , which can enhance the current giving peaks in the current - voltage characteristics at these voltages .
the three - particle current @xmath243 has a non - resonant value of order @xmath244 .
however , @xmath243 is enhanced to order @xmath196 when the energy of one of the two andreev reflections coincides with a bound state energy . for the applied voltage equal to the difference between two bound state energies , @xmath245 ,
two resonances occur simultaneously , i.e. form a resonance consisting of two overlapping single resonances ; see the inset in fig .
[ i3d1plots ] .
this will enhance the current to order @xmath2 close to this voltage , giving a peak in the ivc .
the number of peaks is equal to the number of bound state pairs .
the peaks are located in the voltage interval @xmath246 ; we note that the peak positions are weakly dependent on temperature . to evaluate the height and the width of these peaks , we study the contribution from overlapping resonances at @xmath247 and at @xmath248 .
close to these energies , the phases @xmath249 and @xmath250 , defined in eq .
( [ resphase ] ) , are close to zero , and we find the current spectral density for injection of a quasiparticle from the left ( see appendix [ threepartres ] ) , @xmath251^{(km),l}=\frac { d^{3 } n^{l}(e)n^{r}(e_{3 } ) } { \left|d-4\varphi_{1}^{(k)}\varphi_{2}^{(m ) } + id\left(\varphi_{1}^{(k)}n^{r}(e_{3})+\varphi_{2}^{(m)}n^{l}(e ) \right)\right|^{2 } } .\ ] ] we now expand @xmath252 , @xmath253 in the deviation from perfect overlap in energy , @xmath254 , and in voltage , @xmath255 , and find , using @xmath0 , from eq .
( [ ip3res ] ) @xmath256^{(km),l } = \frac{d \gamma_{0}^{(k ) } \gamma_{3}^{(m ) } } { \left(\delta e_{+}\delta e_{-}/\delta^{2 } - d/4\eta^{(k ) } \eta^{(m ) } \right)^{2}+\lambda^{2}},\ ] ] where @xmath257 , @xmath258 . the energy dependence of the current in eq .
( [ doublebreit ] ) has the form of two resonant peaks with width @xmath259 split by the energy interval @xmath260 at @xmath261 , the peak splitting increasing with increasing @xmath262 .
after integration over energy , the overlapping resonances give a current contribution in the form of a current peak ( @xmath263 ) , @xmath264 in this equation , the densities of states @xmath171 are taken at the entrance and exit energies , @xmath265 and @xmath266 , and the temperature is taken to be zero .
a factor of @xmath267 has been included in eq .
( [ i3res ] ) to take into account the similar resonant process for injection from the right , where @xmath268 and @xmath269 .
the curve for the three - particle current versus voltage thus consists of peaks with heights of order @xmath2 and half - width @xmath270 on top of a background of order @xmath196 .
the background current increases with voltage in the interval @xmath271 as more single resonances come into the integration region . in the interval @xmath272 , the background current decreases due to broadening of the resonances because of leakage associated with incomplete andreev reflection outside the gap . in long symmetric junctions the current peaks form an interesting triangular pattern .
to see this , we first note that if the bound state spectrum were perfectly linear , several of the peaks described by eq .
( [ ip3res])-([i3res ] ) will be situated at the same voltage since @xmath273 ( see also the inset in fig . [ i3long ] ) , and thus the total number of peaks will be reduced while their respective height will be increased .
since the bound state spectrum is not linear , the peaks show splitting .
however , the deviation from linearity is small and in practice the peaks form clusters , giving combined peaks with height roughly equal to the number of clustering peaks . in the interval @xmath271 , the number of peaks in a cluster increases in steps of three from @xmath274 to @xmath275 , etc .
, up to the number of bound states . in the interval
@xmath214 the number of peaks in a cluster decreases in steps of one .
this gives an appearance of a `` peak triangle '' for very long junctions , shown in fig .
[ i3long ] .
this `` peak triangle '' is further enhanced by the background current , which has a similar triangular form , as explained above .
the four - particle current has a non - resonant value of order @xmath276 , which is enhanced to order @xmath244 when the energy of one of the three andreev reflections coincides with a bound state energy .
similar to the three - particle current , overlapping resonances can enhance the magnitude of the current @xmath277 to the order @xmath2 for those voltages where both the first and the third andreev reflections coincide with the bound states , as shown in the upper inset in fig . [ i4d1plots ] .
indeed , it is clear from the explicit form of @xmath278 that when @xmath279 and @xmath280 , then @xmath281 , i.e. the transparency of the mar trajectory is enhanced to unity .
other combinations of the resonances , e.g. when the first and the second andreev reflection occur at bound state energies , will produce peaks of order @xmath196 or smaller , as described in appendix [ fourpartres ] . focusing on the double resonances which produce large ( @xmath282 ) current peaks
, we find that in short junctions with just one pair of bound states , @xmath283 , the double resonance will occur at voltage @xmath284 , provided the energy of the bound state is within the interval @xmath285 .
the spectral density of the current has a form similar to the one in eq .
( [ doublebreit ] ) , the major difference being the small peak splitting@xcite , @xmath286 .
the height of the resulting current peak ( @xmath263 ) is @xmath287_{max}=\frac{\pi de\delta}{h } \frac{1-\left[a(2e^{(0)})\right]^{4 } } { 1+\left[a(2e^{(0)})\right]^{4 } } , \ ] ] where @xmath288 is the andreev reflection amplitude at energy @xmath289 . for longer junctions ,
there are many possibilities to have overlapping resonances .
two bound states at one side of the junction with energies @xmath290 and @xmath291 can give a peak in @xmath277 if @xmath292 . although the height of all peaks is roughly proportional to @xmath2 , numerically the heights ( and widths ) of the peaks may vary considerably depending on the position of the second andreev reflection .
if the second andreev reflection does not occur at the energy of a bound state , the situation is similar to the one described above ; see lower inset in fig .
[ i4d1plots ] .
however , if a bound state is close to the energy of the second andreev reflection , then the current spectral density @xmath293 consists of the three full - transmission peaks with widths @xmath286 which are split within the interval @xmath294 ( triple resonance ) .
the triple resonance has larger spectral weight compared to the double resonance , which results in the larger height and width of the current peak . rigorously speaking
, a triple resonance can only occur in asymmetric junctions because it requires equal distance between neighboring resonances , while the bound state spectrum in symmetric junctions is not equidistant .
however , in long junctions , the deviation from the equidistant spectrum is small , and quasi - triple resonances may therefore occur also in long symmetric junctions .
this effect can be observed in fig .
[ i4d1plots ] , where the four - particle current for a symmetric junction with length @xmath295 consists of three peaks with different heights : the central peak corresponding to the quasi - triple resonance while the two side peaks corresponding to the double resonances with the heights given by eq .
( [ i4max ] ) .
finally , it is worth noting that , similar to the situation for the three - particle current , the peaks will form clusters , giving a smaller number of current peaks than the number of pairs of bound states in long junctions . the studied properties of multiple resonances in three- and four - particle currents allow us to make some general conclusions about resonant behavior of the high order multi - particle currents which determine the total current at small voltage .
the non - resonant magnitude of an @xmath48-particle current is of order @xmath296 at the threshold voltage , @xmath297 , and therefore the total non - resonant current exponentially decreases with the applied voltage ( in transparent junctions , @xmath298 , the current is exponentially small at@xcite @xmath299 ) .
however , multiple resonances may enhance the magnitude of the current by several orders of @xmath2 .
the major question of interest here concerns the maximum value of the resonant current , in particular whether it can be of order @xmath2 at arbitrary small voltage .
to obtain such large current at small voltage , it is necessary to achieve a transmission probability through a high order mar path equal to unity , which implies that the energy of at least every other andreev reflection must coincide with a bound state ( cf .
the discussion in the previous subsection ) . for @xmath300 ,
this means that three or more bound states must be approximately equidistant in energy . since the bound state spectrum is non - equidistant , eq .
( [ resphase ] ) , this is generally not possible if the resonances are narrow ; therefore , in junctions with arbitrary geometry and small transmissivity there are no large current peaks below the voltage @xmath301 .
however , the possibility of a large resonant current exists for junctions with sufficiently large transparency . to find the relevant transparency , let us consider a very long symmetric junction and assume for the moment that the bound state spectrum is equidistant , @xmath302 .
then , from mapping of the @xmath48-th order mar process on a 1d multi - barrier structure ( see fig.[quasi - resonance ] ) , it is clear that if the applied voltage is commensurate with the level spacing , e.g. , @xmath303 , the multi - barrier structure is periodic , and full transmission is achieved leading to a current peak .
this conclusion is valid also for a non - equidistant spectrum if the variation of the interlevel distance does not exceed the width of the full - transmission band .
the deviation of the bound state spectrum from the best linear fit does not exceed the value 0.33@xmath304 , fig.[quasi - resonance ] . on the other hand ,
the width of the full - transmission energy band is @xmath305 for equidistant spectrum and for @xmath306
. thus one should expect large current structures in long symmetric junctions with transparency @xmath307 to occur at voltages @xmath308 . in junctions with smaller transparency
, large current structures may appear only at @xmath309 , as explained before ; see fig .
[ itotl4plots ] .
it is also easy to see that in asymmetric junctions , where the width of the full transmission band for an equidistant spectrum is @xmath286 ( since the relevant resonances at one side of the junction are weakly coupled to each other through the mar process ) , large resonant current at small voltage may exist if @xmath310 .
our numerical investigations confirm that in symmetric junctions when @xmath2 is of the order @xmath311 , the multiple resonances are completely blocked and current peaks are exponentially suppressed at @xmath312 .
adding up the contributions to the current calculated in this paper , we arrive at a rather complex form of current - voltage characteristics ( ivc ) at subgap voltages , as shown in figs .
[ itotd1plots]-[itotl4plots ] . nevertheless , the analysis of the tunnel limit allows us to classify various subgap current structures . here
we will summarize the results of this classification . as a reference system we will take a short ( @xmath183 ) junction where the form of the ivc is well studied @xcite .
the current structures in short junctions can be interpreted as resonant features due to quasi - bound states situated at the edges of the energy gap @xcite , the resonant conditions selecting voltages equal to the gap subharmonics , @xmath313 .
this subharmonic gap structure of the short junction gradually changes with increasing junction length as bound states move down into the gap , giving rise to ivc structures with steps , oscillations and peaks .
major points are : \(i ) the current in the subgap region is considerably enhanced , compared to the short junction case .
this effect is present as soon as the effective length @xmath314 is comparable to , or larger than , the square root of transparency of the junction , @xmath315 .
\(ii ) the main onset of the current in short junctions at @xmath164 shifts downwards in voltage to the value @xmath316 where @xmath317 is the energy of the bound state .
this shift is caused by the resonant two - particle current giving a contribution to the total current of the order of the single - particle current .
\(iii ) for longer junctions , the current onset transforms into a staircase within the voltage interval @xmath214 with the number of steps corresponding to the number of bound states , the step positions being given by @xmath318 .
this is due to the resonances in the two - particle current transported through bound states .
resonant channels open up , one by one , as the voltage increases and bound states enter the `` energy window '' available for two - particle processes . the current plateaus
are not flat but modulated because of oscillations of the density of continuum states .
the period of the modulation is roughly equal to the interlevel distance , and it decreases with the junction length .
the amplitude of the modulation , on the other hand , increases with the junction length .
thus , in long junctions , the current structures take the form of a series of peaks ( see fig .
[ itotd1plots ] ) within the voltage interval @xmath214 .
the position of the peaks has pronounced temperature dependence , scaling with the temperature dependence of the order parameter , while the distance between peaks has a weak temperature dependence .
\(iv ) there is another series of the current peaks whose positions only weakly depend on temperature and are entirely determined by the bound state spectrum : @xmath319 and @xmath320 .
these peaks are caused by the overlap of two resonances in the three- and four - particle currents and they exist in the intervals of applied voltage @xmath246 and @xmath321 respectively .
the heights of these peaks are comparable with the heights of the two - particle current structures ( @xmath282 ) .
\(v ) at voltage smaller than @xmath301 the resonant current structures generally become smaller in magnitude ( at least by one order in @xmath2 ) if the junction transparency is sufficiently small ( @xmath322 ) , and the current decays exponentially when @xmath323 approaches zero ( although for some particular junction lengths there could be huge ( @xmath282 ) current peaks caused by multiple resonances ) . this qualitative difference of the ivc below and above @xmath301
allows one to expect a cross over from power to exponential dependence of ivc in multichannel junctions .
\(vi ) in transparent junctions , all current structures will persist but become smooth ; appreciable current will appear below @xmath301 as soon as @xmath324 .
the current structures completely disappear in fully transparent junctions , @xmath101 , where the ivc does not depend on the junction length ; see fig .
[ itotl4plots ] .
\(vii ) at voltage larger than @xmath325 , the current undergoes oscillations , similar to rowell - mcmillan oscillations @xcite , and the excess current is approached much faster than in short junctions . in low transparency junctions
the excess current is small , @xmath326 , despite strong andreev reflection and large pair current @xmath327 .
we thank e. n. bratus , j. lantz and t. lfwander for discussions . support from nfr and nutek ( sweden ) and from nedo ( japan ) is gratefully acknowledged .
in this appendix , the expansion is derived for the reflection amplitudes in eq .
( [ reflectionexp ] ) for a quasiparticle injected from the left . from the definition of @xmath117 and @xmath328 , eq .
( [ nullvectors ] ) , we know @xmath329 @xmath330 they are related as @xmath331 where @xmath332 and @xmath333 . from this relation , we find @xmath117 in terms of @xmath328 as @xmath334 where @xmath335 .
when @xmath336 , we can make an expansion in this parameter to get to the form @xmath337 similarly we also get @xmath338
in this appendix , we derive the resonant form of the two - particle current , eq .
( [ ip2res ] ) , for a quasiparticle injected from the left .
the definition of @xmath339 is @xmath340 , which using the pseudo - unitarity of the transfer matrices @xmath341 can be written on the form @xmath342 it simplifies in the limit @xmath0 , @xmath343 to @xmath344 inserting the simplified expansion of @xmath339 and the expansion of @xmath345 and @xmath117 from eq .
( [ reflectionexp ] ) into eq .
( [ ipn ] ) , as well as putting @xmath168 , the leakage current density takes the form @xmath346^{l}={\frac{\displaystyle \frac{1- { \left|1-e^{2i\varphi_{0 } } \right|^{2 } } \frac{1-|a_{2}|^{4}}{\left|1-e^{2i\varphi_{2 } } \right|^{2 } } d^{2 } } { \left|2i\varphi_{1}^{(m)}+\displaystyle\frac{d}{2}\left ( \displaystyle\frac{1+e^{2i\varphi_{0}}}{1-e^{2i\varphi_{0}}}+ \displaystyle\frac{1+e^{2i\varphi_{2}}}{1- e^{2i\varphi_{2}}}\right ) \right|^{2}}.}\ ] ] we make an expansion of the phase @xmath347 , where @xmath348 the two - particle current density now takes a breit - wigner form @xmath346^{l}=\frac{\gamma_{0}^{(m)}\gamma_{2}^{(m ) } } { \displaystyle \left(\frac{\delta e-\delta e^{(m)}}{\delta}\right)^{2}+ \left(\frac{\gamma_{0}^{(m)}+\gamma_{2}^{(m)}}{2 } \right)^{2}}\ ] ] where the tunneling rates are given by @xmath349 where @xmath350 are equal to the dos , eq . ( [ densityofstates ] ) at energy @xmath351 .
the resonance is slightly shifted from @xmath96 with @xmath352
in this appendix , the resonant form of the three - particle current , eq .
( [ ip3res ] ) , is derived .
the @xmath353-matrix , which by definition is @xmath354 can be transformed using eq .
( [ m20 ] ) to @xmath355 which can be written in the form @xmath356 it simplifies in the limit of @xmath0 , @xmath357 and @xmath358 to @xmath359.\ ] ] inserting this form of the @xmath353-matrix and the expansion ( [ reflectionexp ] ) for @xmath117 and @xmath360 into eq .
( [ ipn ] ) , as well as putting @xmath168 , the probability current density for injection of a quasiparticle from the left takes the form @xmath361^{l}=\frac{(1-|a_{0}|^{4})(1- { \left|1-e^{2i\varphi_{0 } } \right|^{2 } \left|1-e^{2i\varphi_{3 } } \right|^{2}\left|q\right|^{2}}\ ] ] @xmath362 where @xmath0 is once again used .
since @xmath363 and @xmath364 and the dos at energies @xmath365 , eq . ( [ densityofstates ] ) , are equal to @xmath366 @xmath367 we arrive at the form @xmath361^{l}=\frac{n^{l}(e)n^{r}(e_{3})d^{3 } } { \left|d-4\varphi_{1}^{(k)}\varphi_{2}^{(m ) } + id\left(\varphi_{1}^{(k)}n^{r}(e_{3})+\varphi_{2}^{(m)}n^{l}(e ) \right)\right|^{2 } } .\ ] ]
in this appendix , we discuss the structure of the resonance in the four - particle current . the matrix @xmath368 can be written as @xmath369 @xmath370+\ ] ] @xmath371 @xmath372.\ ] ] from eq .
( [ mmatris ] ) it is clear that in general @xmath373 . when both @xmath374 and @xmath375 are close to a multiple of @xmath376 , @xmath377 , while close to other double resonances , e.g. when @xmath374 and @xmath378 are close to a multiple of @xmath376 , @xmath379 . | we study coherent multiple andreev reflections in quantum sns junctions of finite length and arbitrary transparency .
the presence of superconducting bound states in these junctions gives rise to great enhancement of the subgap current .
the effect is most pronounced in low - transparency junctions , @xmath0 , and in the interval of applied voltage @xmath1 , where the amplitude of the current structures is proportional to the first power of the junction transparency @xmath2 .
the resonant current structures consist of steps and oscillations of the two - particle current and also of multiparticle resonance peaks .
the positions of the two - particle current structures have pronounced temperature dependence which scales with @xmath3 , while the positions of the multiparticle resonances have weak temperature dependence , being mostly determined by the junction geometry . despite the large resonant two - particle current ,
the excess current at large voltage is small and proportional to @xmath4 .
+ pacs : 74.50.+r , 74.80.fp , 74.20.fg , 73.23.ad |
the concept of nanometer - scale regions of polarization , randomly embedded within a non - polar cubic matrix , has become central to attempts to explain the remarkable physical properties of relaxors such as pbmg@xmath2nb@xmath3o@xmath4 ( pmn ) and pbzn@xmath2nb@xmath3o@xmath4 ( pzn ) .
@xcite the existence of these so - called `` polar nanoregions '' ( pnr ) was first inferred from the optic index of refraction studies of burns and dacol on pmn , pzn , and other related systems , @xcite and later confirmed using many different experimental techniques including x - ray and neutron diffraction , @xcite @xmath16pb nmr , @xcite and piezoresponse force microscopy .
@xcite early small - angle x - ray scattering and neutron pair distribution function ( pdf ) measurements on pmn by egami _
et al_. cast doubt on the nano - domain model of relaxors .
@xcite however , the recent pdf analysis of jeong _ et al_. , which shows the formation of polar ionic shifts in pmn below @xmath17k , and which occupy only one third of the total sample volume at low temperatures , provides convincing support for the existence of pnr .
@xcite neutron inelastic scattering data published by naberezhnov _ et al_. on pmn offered the first dynamical evidence of pnr in the form of a prominent broadening of the transverse acoustic ( ta ) mode that coincides with the onset of strong diffuse scattering at 650k , @xcite roughly the same temperature ( @xmath18k ) as that reported by burns and dacol in their optical study of pmn .
this temperature , commonly known as the burns temperature , and denoted by @xmath9 in the original paper , is widely viewed as that at which static pnr first appear .
likely condensing from a soft to mode , distinctive butterfly - shaped and ellipsoidal diffuse scattering intensity contours centered on @xmath7 and @xmath19 bragg peaks , respectively , are seen below @xmath9 in both pmn @xcite and pzn .
@xcite similar diffuse scattering patterns were subsequently observed in solid solutions of pmn and pzn with pbtio@xmath4 ( pmn-@xmath20pt and pzn-@xmath20pt ) ; however these patterns appear only on the ti - poor ( relaxor ) side of the well - known morphotropic phase boundary ( mpb ) .
@xcite the polar nature of the strong diffuse scattering , and thus its association with the formation of pnr , was unambiguously established by several electric field studies of pmn @xcite and pzn-8%pt , @xcite all of which showed dramatic changes in the shape and intensity of the diffuse scattering as a function of field strength and field orientation .
the burns temperature @xmath9 thus represents what is arguably the most important temperature scale in relaxors and is several hundred degrees kelvin higher than the critical temperature @xmath21 ( which for pmn is @xmath22k , but defined only in non - zero electric field ) .
top - temperature dependence of the ta phonon energy width @xmath23 measured with thermal neutrons by wakimoto _
et al_. @xcite middle - diffuse scattering intensity contours measured at ( 110 ) below ( 300k ) and just above ( 650k ) the burns temperature @xmath9 using cold neutrons by hiraka _
et al_. @xcite bottom - diffuse scattering temperature dependence measured with cold neutrons ( open circles ) and thermal neutrons ( dashed line ) by hiraka _
et al_. @xcite , width=288 ] recent studies of the neutron diffuse scattering in pmn by stock _
et al_. , @xcite and especially those by hlinka _ et al_. @xcite and gvasaliya _ et al_. ,
@xcite have proven to be extremely important because they demonstrated the utility of cold neutron spectroscopy to the study of relaxors .
generally speaking cold neutrons are ill - suited to lattice dynamical studies because the longer wavelengths necessarily limit their use to the study of comparatively fewer ( lower-@xmath24 ) brillouin zones . on the other hand cold neutron spectrometers provide significantly better energy ( @xmath25 ) and wave vector ( @xmath14 ) resolution than do their thermal neutron counterparts .
in addition , both the ( 110 ) and the ( 100 ) brillouin zones of pmn are accessible using cold neutrons with wavelengths @xmath26 . the combined cold and thermal neutron study of pmn by hiraka _
et al_. exploited this fact and uncovered several major new findings , two of which are summarized in fig .
[ figure1 ] .
@xcite the first finding is that the temperature at which the strong diffuse scattering vanishes depends on whether the measurement is made with cold or thermal neutrons , i. e. the value of @xmath9 depends on the energy resolution of the spectrometer .
this fact , which is illustrated in the bottom panel of fig .
[ figure1 ] , indicates that the diffuse scattering in pmn , and most likely other relaxors , contains a substantial quasielastic component .
however , a consensus on this issue is lacking ; while the finding of hiraka _ et al_. is consistent with the study of gvasaliya _
et al_. , @xcite it contradicts that of hlinka _
et al_. @xcite a second major finding is displayed in the middle panel of fig .
[ figure1 ] where intensity contours of the diffuse scattering measured with cold neutrons around ( 110 ) are shown to exhibit markedly different reciprocal space geometries at 300k and 650k .
these data unambiguously demonstrate the presence of two distinct types of diffuse scattering in pmn . as the pnr
are absent at 650k @xmath27 , the physical origin of the bow - tie - shaped diffuse scattering cross section is believed to come primarily from the underlying chemical short - range order .
we shall refer to this as csro , although another commonly - used term for this is compositionally / chemically ordered regions ( cor ) .
@xcite an intriguing perspective on the burns temperature in pmn is provided in the top panel of fig .
[ figure1 ] where the ta phonon energy width @xmath23 , which is inversely proportional to the phonon lifetime , is plotted as a function of temperature for pmn at the scattering vector @xmath28 .
these data were measured with thermal neutrons by wakimoto _
et al_. @xcite and are consistent with those of naberezhnov _
et al_. in that the ta mode begins to broaden at @xmath29k , the same temperature where the strong diffuse scattering first appears .
these data also show that the ta broadening achieves a maximum value ( minimum lifetime ) near 420k , which coincides with the value of @xmath9 measured with cold neutrons .
these data raise the question of how to interpret the burns temperature @xmath9 properly in that they paint a picture , markedly different from that currently held , in which the pnr are dynamic below @xmath30k and condense into static entities only at a much lower temperature of 420k .
the goal of our study then is to determine the intrinsic value of the burns temperature @xmath9 and to clarify its relationship to the diffuse scattering , lattice dynamics , and structure in pmn . to this end
we have carried out extensive measurements of the neutron diffuse scattering cross section from 10k to 900k , far below @xmath21 and well above the nominal value of @xmath9 , that probe timescales from @xmath0s to @xmath1s .
we find that a 300-fold improvement in energy resolution over that used by hiraka _
et al_. , @xcite obtained using neutron backscattering techniques reproduces the same onset temperature for the diffuse scattering ; hence the intrinsic burns temperature @xmath9 for pmn is 420k .
at the same time an enormous change in the thermal expansion is observed near 420k , which is indistinguishable from zero at low temperature .
given the revised value of @xmath9 , this result implies the existence of a direct influence of the pnr on the intrinsic structural properties of pmn .
we also present new data on the effects of the pnr on the lattice dynamics through measurements of the temperature and wave vector dependence of the long - wavelength ta phonon energy width @xmath23 measured near ( 110 ) from 25k to 900k .
we find that ta modes with reduced wave vectors @xmath31@xmath12 exhibit the same energy width at all temperature whereas those with @xmath32@xmath12 exhibit a strongly temperature - dependent broadening similar to that shown in the top panel of fig.[figure1 ] .
this behavior contrasts with that observed in thermal neutron studies of the to mode , which exhibits a broadening for all @xmath33@xmath12 .
previous neutron scattering work on pmn-60%pt by stock _
et al . _ , @xcite a material in which there is no strong , temperature dependent diffuse scattering , and thus no polar nanoregions , found no evidence of any ta phonon broadening . in this context ,
our data lend extremely strong support to the pnr model : the lifetimes of ta modes with wavelengths comparable in size to the pnr are strongly diminished by the pnr , whereas long - wavelength ( low @xmath14 ) ta phonons simply average over the pnr and are unaffected .
the neutron scattering data presented here were obtained using the bt9 thermal neutron triple - axis spectrometer , the spins cold neutron triple - axis spectrometer , and the cold neutron high flux backscattering spectrometer ( hfbs ) , all of which are located at the nist center for neutron research ( ncnr ) . on bt9 , measurements of the phonons and diffuse scattering were made at a fixed final ( thermal ) neutron energy @xmath3414.7mev ( @xmath35 ) using the ( 002 ) bragg reflection of highly - oriented pyrolytic graphite ( hopg ) crystals to monochromate and analyze the incident and scattered neutron beams , respectively .
horizontal beam collimations were 40@xmath36 - 47@xmath36-s-40@xmath36 - 80@xmath36 ( s = sample ) . a special
, non - standard high @xmath14-resolution configuration was employed to measure the thermal expansion in which the ( 004 ) bragg reflection from a perfect ge crystal was used as analyzer and horizontal beam collimations were tightened to 15@xmath36 - 47@xmath36-s-20@xmath36 - 40@xmath36 .
the choice of ge was motivated by the close matching between the pmn ( 022 ) ( 1.431 ) and ge ( 004 ) ( 1.414 ) @xmath37-spacings , which provides a significant improvement in the instrumental @xmath14-resolution .
@xcite on spins , which sits on the cold neutron guide ng5 , the phonon and diffuse scattering measurements were made at a fixed final neutron energy @xmath38mev ( @xmath39 ) also using the ( 002 ) bragg reflection of hopg crystals as monochromator and analyzer .
a liquid - nitrogen cooled be filter was located after the sample to remove higher order neutron wavelengths from the scattered beam , and horizontal beam collimations were set to guide-80@xmath36-s-80@xmath36 - 80@xmath36 .
the resultant elastic ( @xmath40 ) energy resolution for the spins measurements was @xmath41mev half - width at half - maximum ( hwhm ) .
the high - flux backscattering spectrometer was used to look for dynamics that might be associated with the strong diffuse scattering below @xmath9 .
this instrument uses a mechanically - driven si(111 ) monochromator to doppler shift the energies of incident neutrons over a narrow range centered about 2.08mev .
neutrons are backscattered from the monochromator and proceed towards the sample where they are scattered into a 12m@xmath42 array of si ( 111 ) crystals that serve as analyzer .
these neutrons are then backscattered a second time by the analyzer , which selects the final neutron energy @xmath43mev , into a series of detectors positioned about the sample .
the effective angular acceptance of each detector is @xmath44 .
the hfbs instrument is described in further detail elsewhere .
@xcite the elastic energy resolution for the hfbs measurements described here was @xmath45@xmath46ev ( hwhm ) .
two high - quality single crystals of pmn , labeled pmn#4 and pmn#5 , were used in this study ; both were grown using a top - seeded solution growth technique .
@xcite the crystal growth conditions were determined from the pseudo - binary phase diagram established for pmn and pbo .
the pmn#4 and # 5 crystals weigh 2.7 g ( 0.33@xmath47 ) and 4.8 g ( 0.59@xmath47 ) , respectively . at 300k the mosaic of each crystal measured at ( 220 ) is less than 0.04@xmath48 full - width at half - maximum ( fwhm ) .
loss of pbo , the formation of a pyrochlore phase , and the reduction of nb@xmath49 are known to occur in pmn single crystals when subjected to high temperatures under vacuum for extended periods of time .
this process results in a dramatic blackening of the crystal , which is normally of a transparent gold / amber color .
while dielectric measurements on such darkened crystals reportedly show little difference from those on unheated samples , @xcite our measurements reveal a diminishment of the diffuse scattering intensity after sustained and repeated heating .
therefore experiments on the larger pmn crystal # 5 were limited to 600k or less , while pmn crystal # 4 was used to obtain data above 600k .
both samples were mounted with an [ 001 ] axis oriented vertically , giving access to reflections of the form @xmath50 .
for the high temperature experiments , pmn crystal # 4 was wrapped in quartz wool , mounted in a niobium holder secured by tungsten wire , and then loaded into a water - cooled furnace capable of reaching temperatures from 300k to 1800k .
pmn crystal # 5 was mounted onto an aluminum sample holder assembly placed inside an aluminum sample can , and then loaded inside the vacuum space of a closed - cycle @xmath51he refrigerator that provides access to temperatures from 10k to 700k .
each sample has a cubic lattice spacing of @xmath52at 300k ; thus 1rlu ( reciprocal lattice unit ) equals @xmath53@xmath12 .
diffuse scattering intensity contours in pmn measured on spins near ( 100 ) are shown using a linear gray scale at ( a ) 300k , ( b ) 650k , and ( c ) 900k .
diffuse scattering intensity contours measured on bt9 at 900k near ( 300 ) using thermal neutrons are shown in ( d ) .
the data shown in panels ( a ) and ( b ) are from hiraka _
et al_. @xcite schematic diagrams of the diffuse scattering intensity contours below ( low-@xmath54 ) and at @xmath9 ( high-@xmath54 ) are shown in the upper right.,width=288 ] contour plot of the difference in diffuse scattering intensity near ( 110 ) measured at 550k and 400k .
the result is the characteristic figure `` 8 '' pattern , observed previously by vakhrushev _
et al_. @xcite , width=288 ] the diffuse scattering in pmn crystal # 5 was studied by hiraka _
et al_. @xcite at and below the nominal burns temperature @xmath9 with good energy resolution ( @xmath55mev hwhm ) using cold neutrons .
data were measured in the ( 100 ) and ( 110 ) brillouin zones and are represented schematically in the upper right portion of fig .
[ diffuse_contours ] .
the horizontal and vertical axes in this figure correspond to the components of the scattering vector @xmath56 , which are measured in reciprocal lattice units ( rlu ) .
the `` low-@xmath54 '' ( @xmath57 ) regime is dominated by contributions from the pnr , where the diffuse scattering intensity contours near ( 100 ) resemble a butterfly and those near ( 110 ) resemble an ellipse for which the long axis is oriented perpendicular to @xmath58 .
this is shown in fig .
[ diffuse_contours](a ) where diffuse scattering data near ( 100 ) at 300k reveal an intense butterfly - shaped pattern .
@xcite we note here that we see no evidence of any transverse component " to the diffuse scattering in any of our pmn crystals like that reported previously by vakhrushev _
et al_. @xcite we assume that this component was due to the crystal imperfections mentioned by these authors in their pmn sample , which gave rise to a powder ring in their data at ( 100 ) .
the same butterfly / ellipsoidal diffuse scattering geometry was shown to persist in single crystals of pmn-@xmath20pt and pzn-@xmath20pt in studies by , respectively , matsuura _
et al_. and xu _ et al_. for compositions spanning the ti - poor ( relaxor ) side of the morphotropic phase boundary ( mpb ) .
@xcite these results also completely refute those of la - orauttapong _ et al_. who reported that the orientation of the strong diffuse scattering varies with ti content in pzn-@xmath20pt and concluded that the pnr orientation changes with doping .
@xcite for ti - rich ( tetragonal ) pmn-@xmath20pt compositions just beyond the mpb , matsuura _
et al_. found that the strong , temperature dependent diffuse scattering vanishes and is replaced by critical scattering .
@xcite matsuura _
et al_. also found that the @xmath14-integrated diffuse scattering intensity increases with ti content on the ti - poor side of the mpb , peaks near the mpb , then drops dramatically on crossing the mpb .
this finding is significant because it suggests that an intriguing and direct correlation exists between the pnr and the piezoelectric coefficient @xmath59 , which exhibits the same dependence on ti content .
@xcite a model based on pancake - shaped , ferroelectric domains has been used successfully to fit the three - dimensional diffuse scattering distributions measured in pzn-@xmath20pt with high - energy x - rays .
@xcite a similar type of real - space structure has been proposed to explain the diffuse scattering in the relaxor kli@xmath60ta@xmath61o@xmath4 .
@xcite on the other hand , alternative models explaining the same diffuse scattering distributions have also been proposed .
@xcite in the `` high-@xmath54 '' regime ( @xmath62 ) there are no pnr , and the associated butterfly - shaped diffuse scattering is no longer present . the weak diffuse scattering that remains is thus argued to originate primarily from the underlying chemical short range order ( csro ) , which reflects weak correlations between the mg@xmath63 and nb@xmath49 cations on the @xmath64-site of the perovskite @xmath65o@xmath4 structure . in this regime
the shapes of the diffuse scattering contours are radically different , resembling a bow - tie in both @xmath66 and @xmath67 brillouin zones in which the diffuse scattering extends mainly parallel to @xmath58 . the only difference between the contours near ( 100 ) and
( 110 ) appears to be in the orientation of the triangular regions of diffuse scattering , which point in towards ( 100 ) , but away from ( 110 ) .
data taken near ( 100 ) at 650k are displayed in fig .
[ diffuse_contours](b ) .
@xcite at this temperature the diffuse scattering intensities , shown using a linear gray scale , are much weaker than those of the butterfly pattern at 300k .
we further note that the intensities increase ( become darker ) from left to right in panels ( b ) and ( c ) , which corresponds to increasing @xmath24 . given the @xmath68 dependence of the neutron diffuse scattering cross section , this intensity signature implies the presence of some short - range , correlated displacements @xmath69 since otherwise , if @xmath70 , there would be no @xmath24-dependence .
thus the weak diffuse scattering is not solely due to csro . in fig .
[ subtraction_figure ] we plot the difference between the diffuse scattering intensities measured near @xmath71 at 400k and 550k .
the resulting contours approximately reproduce the `` figure 8 '' pattern observed previously by vakhrushev _
et al_. , @xcite , and indicates that the ellipsoidal diffuse scattering has a strong temperature dependence .
by contrast , the bow - tie - shaped diffuse scattering is effectively subtracted out in this analysis , which confirms that it has little to no temperature dependence .
thus the high - temperature diffuse scattering is not associated with the burns temperature @xmath9 or with the formation of long - range polar correlations below @xmath21 .
we note that such a strictly incoherent treatment of the high - temperature ( csro ) and low - temperature ( pnr ) components of the total diffuse scattering , as described here , ignores the inevitable cross terms that must exist between them . on the other hand ,
if the high - temperature scattering does arise primarily from csro , then it should be largely independent of the ionic displacements that give rise to the butterfly - shaped diffuse scattering ( pnr ) below @xmath9 .
the relative weakness of the high - temperature diffuse scattering compared to that at low temperatures also suggests such cross terms should be weak , and this appears to be supported by the simple subtraction analysis presented in fig .
[ subtraction_figure ] in that one effectively recovers the ellipsoidal ( not bow - tie ) intensity contours . for this reason we believe it is a reasonable first approximation to treat the two diffuse scattering components as being nearly independent .
diffuse scattering intensity measured on bt9 at 300k as a function of reduced momentum transfer relative to ( 100 ) ( open circles ) and ( 200 ) ( solid circles ) .
the dotted lines isolate the diffuse scattering component from the bragg peak .
the inset shows the orientation of the scan relative to the familiar butterfly pattern measured near ( 200 ) at 100k.,width=288 ] diffuse scattering contours measured on bt9 near ( 200 ) are shown at ( a ) 100k , ( b ) 300k , and ( c ) 600k.,width=288 ] we extended the diffuse scattering measurements to temperatures well above @xmath9 using pmn crystal # 4 , which we reserved for very high temperature experiments .
previous data taken on pmn and pzn crystals heated to 1000k revealed significant evidence of sample decomposition , so we limited our measurements to 900k .
diffuse scattering intensity contours at 900k for the ( 100 ) and the ( 300 ) brillouin zones are presented in fig .
[ diffuse_contours](c ) and [ diffuse_contours](d ) , respectively .
although the diffuse scattering near ( 100 ) is quite weak , it is consistent with the bow - tie geometry observed in fig .
[ diffuse_contours](b ) at 650k .
we exploited the @xmath72 dependence of the neutron diffuse scattering cross section to obtain higher intensity by using thermal neutrons to access ( 300 ) .
the ( 300 ) diffuse scattering intensity contours are shown in fig .
[ diffuse_contours](d ) , where the bow - tie pattern observed at 650k is still present at 900k .
the contours are truncated for @xmath73rlu because of a mechanical limit on the maximum scattering angle available on bt9 , but the triangular region on the low-@xmath24 side of ( 300 ) is clearly evident . that the bow - tie - shaped diffuse scattering persists to such high temperature provides what is perhaps the most convincing evidence that it arises mainly from csro . during the course of our measurements we noticed that the diffuse scattering in pmn crystal # 4 had diminished and was noticeably weaker than that in pmn crystal # 5 , which we had never exposed to temperatures above 650k .
we also observed a broad ring of scattering passing directly through @xmath74 , which is shown in fig .
[ diffuse_contours](d ) .
this feature was never observed prior to heating this crystal to 900k .
given the length of time spent at high temperatures , this feature most likely corresponds to a powder ring arising from partially decomposed regions of pmn crystal # 4 , which had turned entirely black after exposure to high temperatures .
these regions do not affect any other data presented in this paper because the 900k measurements on pmn crystal # 4 were the last ones performed on this sample .
therefore the powder ring appears only in fig .
[ diffuse_contours](d ) .
it is instructive to compare these results with those on pmn-60%pt , a composition that lies well beyond the morphotropic phase boundary and undergoes a first - order ferroelectric transition from a cubic to a tetragonal phase near 540k .
an extensive study of this material using neutron and high - energy x - ray scattering methods found no sign of the strong , butterfly - shaped diffuse scattering at low temperatures .
@xcite this result lends further support to our association of the strong , temperature dependent diffuse scattering with the pnr , which are absent in pmn-60%pt .
neutron measurements on pmn-60%pt do , however , reveal the presence of bow - tie shaped diffuse scattering intensity contours at all temperatures studied , which supports the identification of such diffuse scattering with chemical short - range order between cations on the @xmath64 site of the pmn perovskite @xmath65o@xmath4 structure .
this picture is supported by theoretical work @xcite as well as @xmath75nb nmr , @xcite electron microscopy , @xcite and polarized raman scattering @xcite measurements .
all of these studies suggest that there is no temperature dependence to the bow - tie shaped diffuse scattering below @xmath76k , which is consistent with our results on pmn over the extended temperature range . during our study of pmn
we discovered that the diffuse scattering near ( 200 ) is not as weak as previously believed .
@xcite to confirm this finding , we made detailed measurements of the diffuse scattering intensity near ( 200 ) at 300k and 100k along a trajectory in reciprocal space that follows one wing of the butterfly intensity contour ; this is shown by the dashed line in the inset to fig .
[ diffuse200 ] .
the results of the 300k scan are compared to an identical scan measured in the ( 100 ) zone , both of which are shown in fig .
[ diffuse200 ] .
these data demonstrate that the ( 100 ) diffuse scattering cross section , represented by the dotted lines passing through the open circles , is substantially larger than that at ( 200 ) , designated by the solid circles .
this result supports the model of hirota _
et al_. in which the ( unexpectedly ) weak ( 200 ) diffuse scattering cross section observed in pmn and other relaxors can be explained by the presence of a uniform shift or displacement of the pnr relative to the non - polar cubic matrix along the direction of the local pnr polarization .
@xcite indirect evidence for the existence of this shift has been obtained from neutron scattering measurements of the anisotropic response of the diffuse scattering in pzn-8%pt to an electric field applied along the [ 001 ] direction .
@xcite fig .
[ diffuse200_temp ] shows diffuse scattering intensity contours measured on bt9 at 100k , 300k , and 600k near(200 ) ; these data illustrate that the ( 200 ) diffuse scattering intensity follows the same temperature dependence as that measured in other brillouin zones , where the diffuse scattering is much stronger .
as the temperature is raised the diffuse scattering intensity decreases in the same manner as that previously measured and observed in the ( 100 ) , ( 110 ) , and ( 300 ) brillouin zones .
this proves that the diffuse scattering measured at ( 200 ) has the same origin as that in other zones , i. e. that it is associated with the formation of pnr . at 600k in panel ( c ) one can already see the emergence of the bow - tie - shaped diffuse scattering that is otherwise obscured by the stronger pnr - related diffuse scattering at lower temperatures .
these data are important because they support the mode - coupling analysis of stock _
et al_. @xcite , which assumes that the diffuse scattering in pmn in the ( 200 ) and ( 220 ) brillouin zones is much weaker than that in the ( 110 ) zone .
thus we emphasize that while the neutron diffuse scattering cross - section near ( 200 ) is not zero , it is small and consistent with previous structure factor calculations .
diffuse scattering intensity contours in pmn measured on the ncnr disk chopper spectrometer ( dcs ) near ( 100 ) are plotted on a logarithmic gray scale at 300k .
open circles represent the reciprocal space locations of the detectors of the high flux backscattering spectrometer relative to the butterfly - shaped intensity contours .
each open circle is a separate detector , and the detector numbers are indicated .
these dcs data are taken from xu _
et al_. @xcite , width=360 ] ( color online ) ( a ) contour plot of the diffuse scattering intensity measured on the hfbs as a function of energy transfer and temperature .
contours are shown on a linear intensity scale ; dashed lines indicate the full - width at half - maximum ( fwhm ) of the peak linewidth at each temperature .
data were summed over detectors 10 - 16 as illustrated in fig .
[ hfbs_det ] .
panels ( b ) , ( c ) , and ( d ) show inelastic scans at 700k , 300k , and 100k . the horizontal bar in panel ( c ) represents the instrumental fwhm elastic energy resolution ( @xmath77).,width=360 ] the reciprocal space geometry of the strong diffuse scattering in pmn was first characterized using x - ray diffraction and is consistent with the neutron scattering data we have presented here .
@xcite the energy resolution provided by x - ray diffraction ( @xmath78mev ) is typically much broader than that of thermal neutrons ( @xmath79mev ) ; thus it was assumed that the strong diffuse x - ray scattering originated from low - energy , soft , transverse optic ( to ) phonons that were captured by the large energy resolution .
@xcite however the cold neutron data of hiraka _
provide a much narrower elastic energy resolution of @xmath80mev hwhm and show , unambiguously , that the diffuse scattering cross section contains a component that is static on timescales of at least @xmath81ps below 420k as illustrated in fig .
[ figure1 ] .
this result was subsequently confirmed on a separate pmn crystal by the neutron study of gvasaliya _ et al_. which employed comparable energy resolution .
@xcite hence the observed strong diffuse scattering can not simply be the result of a soft , low - lying to phonon .
the to mode must condense and/or broaden sufficiently to produce the elastic diffuse scattering cross section observed by hiraka _
et al_. such a scenario is in fact suggested by the corresponding thermal neutron data taken on bt9 using a somewhat coarser energy resolution of @xmath82mev hwhm . as shown in fig .
[ figure1 ] , an apparent elastic diffuse scattering cross section is observed up to temperatures as high as 650k .
( a ) temperature dependence of the diffuse scattering measured on the hfbs ; data are integrated in energy from @xmath83 5 @xmath84 and in @xmath24 over detectors 10 - 16 as illustrated in fig .
[ hfbs_det ] .
( b ) temperature dependence of the diffuse scattering intensity measured on spins at ( 1.05,0.95,0 ) by hiraka _
et al_. @xcite the spins energy resolution of @xmath85=120@xmath46ev hwhm provided the energy integration.,width=312 ] motivated by these results , we looked for evidence of a dynamic component of the strong diffuse scattering using the ncnr high - flux backscattering spectrometer ( hfbs ) , which provides an elastic energy resolution @xmath45@xmath46ev hwhm .
we oriented pmn crystal # 4 in the @xmath50 scattering plane , which is the same geometry used for the triple - axis studies discussed in the previous sections .
[ hfbs_det ] displays the resulting locations of each of the 16 hfbs detectors relative to the butterfly - shaped diffuse scattering pattern at ( 100 ) measured previously using the ncnr disk chopper spectrometer ( dcs ) .
@xcite from this figure it can be seen that detectors 10 through 16 sample different parts of the wings of the strong diffuse scattering .
in particular , detectors 13 and 14 lie close to the ( 100 ) bragg peak .
because the instrumental @xmath24-resolution of the hfbs is relatively poor , we checked for the presence of bragg contamination in the integration analysis by removing contributions from detectors 13 and 14 .
this did not change any of the results . moreover
the pmn study by wakimoto _
et al_. showed the ( 202 ) bragg peak intensity changes by less than 10% between 50k and 300k .
@xcite therefore , in the following analysis we have integrated the intensity from detectors 10 through 16 .
the energy dependence of the diffuse scattering as a function of temperature is illustrated in fig .
[ contour_figure ] .
panel ( a ) shows a color contour plot of the peak linewidth in energy as a function of temperature after subtraction of a high temperature background .
this background was taken to be the average of the intensity measured at 600k and 700k because no strong diffuse scattering is observed with either cold neutron triple - axis or neutron backscattering methods at these temperatures .
the distance between the dashed lines represents the measured energy width ( full - width at half - maximum ) as a function of temperature .
typical energy scans are displayed in panels ( b ) , ( c ) , and ( d ) at 700k , 300k , and 100k , respectively .
we see that the peak centered at @xmath86 is equal to the instrumental energy resolution @xmath45@xmath46ev hwhm at all temperatures . based on this analysis
, we conclude that the strong diffuse scattering is elastic on timescales of at least @xmath87ns .
our results are thus consistent with the neutron - spin echo study on pmn of vakhrushev _
et al_. @xcite fig .
[ integrated](a ) shows the temperature dependence of the diffuse scattering intensity integrated over @xmath88@xmath46ev .
these data are compared to those measured on pmn crystal # 5 using the cold neutron spectrometer spins , which are plotted in panel ( b ) .
that the temperature dependences from the backscattering and spins measurements agree within error in spite of a 300-fold ( 120@xmath46ev/0.4@xmath46ev ) difference in energy resolution proves that static , short - range polar order first appears at much lower temperatures than has been understood from previous data measured with much coarser energy resolution .
these data thus demand a revised value for the burns temperature : @xmath89k .
we mention , however , that if the diffuse scattering energy width obeys an arrhenius or some power law form , which has been suggested for spin glasses in ref . , then a more detailed analysis based on data taken closer to the onset of the diffuse scattering at 420k will be required to confirm ( or reject ) the existence of such alternative dynamic contributions to the diffuse scattering energy width . to gain a better understanding of the apparent quasielastic nature of the diffuse scattering
, we examined the temperature dependence of the low - energy transverse acoustic ( ta ) modes in greater detail .
in particular we focused on measurements of the temperature dependent broadening of the transverse acoustic phonon over a range of reduced wave vectors @xmath14 that approach the zone center ( @xmath90 ) .
these data are discussed in the following section .
an extensive series of inelastic measurements were made on both pmn crystals # 4 and # 5 in the ( 110 ) brillouin zone using the spins spectrometer in an effort to map out the temperature dependence of the ta mode for reduced wave vectors @xmath91 and @xmath92 , where @xmath93rlu @xmath94@xmath12 is the wave vector associated with the so - called `` waterfall '' effect , and below which the long - wavelength , soft to modes in pmn , pzn , and pzn-8%pt are observed to broaden markedly at temperatures below @xmath9 .
@xcite this zone was chosen because the ta phonon dynamical structure factor for ( 110 ) is much larger than that for ( 100 ) .
the ta phonon energy lineshapes were studied at two very different values of the reduced wave vector @xmath14 , measured relative to the ( 110 ) zone center , for temperatures between 100k and 900k .
these data are presented in fig .
[ ta_spins ] .
the data shown on the left - hand side of this figure correspond to @xmath95 or @xmath96rlu = 0.05rlu = 0.077@xmath12 , which is less than half of @xmath97 .
these data show that the ta lineshape at this small wave vector remains sharp and well - defined at all temperatures and has an intrinsic energy width that is larger than the instrumental resolution ( shown by the small , solid horizontal bar at 500k ) . by contrast , the data on the right - hand side of fig.[ta_spins ] correspond to @xmath98 or @xmath99rlu = 0.14rlu = 0.22@xmath12 , which is nearly equal to @xmath97 . in this case
it is quite evident that the ta lineshape broadens dramatically , becoming increasingly ill - defined below @xmath9 , especially at 300k , but then sharpens at lower temperature , e. g. at 100k .
this behavior is the same as that observed for the soft , zone - center to mode in pmn by wakimoto _
et al_. measured at ( 200 ) .
@xcite ta phonon lineshapes in pmn measured on spins with cold neutrons at @xmath100 and ( 1.1,0.9,0 ) from 100k to 900k .
, width=312 ] in addition to these cold neutron measurements of the ta mode using spins , other measurements were made using thermal neutrons on bt9 to characterize the to mode lineshapes for reduced wave vectors @xmath101 from 300k to 900k . in these experiments data
were taken using a water - cooled furnace with pmn crystal # 4 in both the ( 300 ) and ( 100 ) brillouin zones for @xmath102rlu and are shown in fig.[figure_100_300 ] . as can be seen on the left - hand side of this figure , both the ta and to modes
are well - defined at 900k .
the scattering intensity below 1 - 2mev increases sharply because of the comparatively coarser energy resolution intrinsic to bt9 ( @xmath103mev hwhm ) , which uses thermal neutrons , compared to that on spins .
even so , the ta mode is easily seen at both 900k and 700k . however at 500k the ta mode is less well - defined .
this occurs in part because the mode has broadened , but also because the low - energy scattering has increased , which results from the onset of the strong diffuse scattering due to the pnr . at 300k
the ta mode is nearly indistinguishable from the sloping background , and the to mode has become significantly damped .
one also sees a gradual softening of the to mode from 900k down to 300k that is of order 1 - 2mev .
the same data were taken in the ( 100 ) zone , which appear in the right - hand portion of fig .
[ figure_100_300 ] .
essentially the same trends are observed in this zone , with the ta mode again becoming almost indistinguishable from background at 300k .
one difference is that the ta mode is better separated at all temperatures from the steep increase in scattering seen at low energies .
these data are consistent with the conjecture of stock _
et al_. @xcite that the low - energy ta phonon is coupled to the diffuse scattering centered around the elastic position .
the inset on the top panel on the right - hand side displays the full ta - to phonon spectrum at 900k out to 20mev . ta and to phonon lineshapes in pmn measured on bt9 with thermal neutrons near the waterfall wave vector at @xmath104 and ( 1,-0.14,0 ) from 300k to 900k .
, width=312 ] the square of the phonon energy is plotted versus temperature for the zone - center to modes measured at ( 200 ) ( open diamonds ) and ( 220 ) ( solid diamonds ) , taken from wakimoto _
et al_. and stock _
et al_. @xcite , as well as for three ta modes measured at ( 1.035,0.965,0 ) ( open triangles ) , ( 1.1,0.9,0 ) ( open circles ) , and ( 3,-0.14,0 ) ( solid circles ) .
linear behavior is consistent with that expected for a conventional ferroelectric soft mode.,width=288 ] the square of the energies of both the ta and to modes presented in the two previous figures are plotted versus temperature in fig .
[ energy_squared ] .
this is done to draw attention to the similarity between the low - energy lattice dynamics of pmn and the behavior expected for a conventional ferroelectric , for which the soft mode frequency squared @xmath105 varies linearly with @xmath106 .
data for the zone - center soft mode measured at ( 200 ) and ( 220 ) have been taken from wakimoto _
et al_. and stock _
et al_. and added to the top of fig .
[ energy_squared ] for ease of comparison .
@xcite the data taken by wakimoto _
et al_. are based on energy scans measured at constant-@xmath24 at the zone center , whereas stock _
et al_. determined the zone - center to energies by extrapolating from data obtained at non - zero @xmath14 .
the extrapolation technique permits phonon energies to be extracted in the temperature range where the zone - center to mode is heavily damped .
these values for the zone - center to phonon energy have been confirmed by infrared spectroscopy .
@xcite what this figure immediately reveals , then , is a surprising ferroelectric character of the zone center to mode above @xmath9 and below @xmath21 , and a corresponding change in the behavior of the ta mode energy that is bracketed by the same two temperature scales , which are denoted by the two vertical dash - dotted lines .
we note that no such change in the acoustic phonons was observed in pmn-60%pt where pnr and the associated strong diffuse scattering are absent ; therefore the softening of the ta mode is directly related to the development of pnr .
@xcite a further interesting feature is the minimum that is present in all of these data near 420k , which is the same temperature at which the strong diffuse scattering first appears when measured with high energy resolution , namely the revised value of the burns temperature .
therefore , the onset of the diffuse scattering is directly associated with the softening of the to mode ; this is yet further evidence that it is associated with the formation of static , short - range polar correlations .
it is very important to note that the square of the ta phonon energy measured at @xmath58=(1.1,0.9,0 ) , which corresponds to @xmath107rlu , shows a much more pronounced minimum at 420k than does that measured at @xmath58=(1.035,0.965,0 ) , which corresponds to @xmath108rlu .
this shows that long - wavelength ta phonons exhibit a much weaker response to the formation of static short - range , polar correlations .
this can be understood in terms of a simple physical picture in which those phonons with wavelengths comparable to the size of the pnr are strongly affected ( damped ) by the pnr whereas longer wavelength phonons simply average over the pnr and are thus not affected by the presence of static , short - range polar correlations .
this idea was previously proposed to explain the anomalous damping of the to mode for wave vectors @xmath109 near the zone center .
@xcite however , no strong diffuse scattering is seen in pmn-60%pt and thus no pnr are present , even though the anomalous to broadening is still observed ; hence this to broadening , which gives rise to the waterfall effect , can not be associated with the presence of static , short - range polar correlations . on the other hand ,
the idea that the acoustic phonons are affected by pnr is confirmed by the absence of any such acoustic phonon broadening in pmn-60%pt .
thus pnr have a significant effect on the low - energy acoustic phonons over a limited range of reduced wave vectors that may be related to the size of the pnr . in light of the diffuse and inelastic scattering data that have been analyzed so far
, we now turn to the detailed measurement of the thermal expansion in pmn .
a variety of experimental techniques have been used to measure the thermal expansion in both single crystal and powder samples of pmn , which includes x - ray diffraction , laser dilatometry , and neutron diffraction .
@xcite here we present extremely high @xmath14-resolution neutron measurements ( @xmath110rlu = 0.0008@xmath12 hwhm ) of the cubic unit cell lattice parameter @xmath111 of pmn crystal # 5 using the special triple - axis configuration described in the experimental section .
this configuration , which employs a perfect single crystal of ge as analyzer , also provides an exceptional energy resolution @xmath112@xmath46ev hwhm .
the resulting data are plotted in fig .
[ thermal_expansion ] in terms of the strain @xmath113 , where @xmath114 is the lattice parameter at 200k .
measurements were made on heating from 30k to 580k using a closed - cycle @xmath51he refrigerator after having first cooled to 30k .
x - ray data measured by dkhil _
et al_. @xcite on a single crystal of pmn , shown as open squares , are provided for comparison .
lattice strain of single crystal pmn derived from the ( 220 ) bragg reflection measured from 30k to 580k on bt9 .
data were obtained on heating using a perfect ge(004 ) analyzer , an incident neutron energy @xmath115mev , and horizontal beam collimations of 15@xmath36 - 47@xmath36-s-20@xmath36 - 40@xmath36 to obtain extremely high @xmath14-resolution .
x - ray data from dkhil _
et al_.are indicated by the open squares for comparison .
@xcite.,width=288 ] several features are interesting to note .
first , at low temperature the system exhibits an invar - like effect in which the cubic lattice parameter changes by less than 0.001 ; indeed , the data below 320k are consistent with a thermal expansion of zero . at higher temperatures
, however , the average thermal expansion is @xmath151/k . that these disparate regions of null and high rates of thermal expansion bracket the revised value for the burns temperature suggests that a direct connection exists between the onset of static , short - range polar correlations and the structural properties of pmn .
this behavior seems to be consistent with that of pzn , which also exhibits an increase in the thermal expansion at temperatures above that where the diffuse scattering first appears .
@xcite there is ample evidence of similar behavior reported by other groups in samples of pmn and pmn-@xmath20pt .
a low - temperature invar - like effect was observed in single crystal pmn-10%pt , where a transition to a high rate of thermal expansion was found at 400k ; the thermal expansion for this sample at high temperature is @xmath1161/k , which is very close to that measured here .
@xcite the x - ray study of ceramic samples of pmn by king _
et al_. also show a transition between low and high rates of thermal expansion , but larger values for both .
@xcite x - ray and neutron work conducted by bonneau _
et al_. on pmn powders yielded onset temperatures and values for the thermal expansion consistent with our single crystal measurements .
@xcite finally , an invar effect was also observed in the laser dilatometry study by arndt and schmidt on a ceramic sample of pmn , which covered a range from 300k to 800k .
@xcite even though there is a general trend towards a larger thermal expansion for temperatures above that where the diffuse scattering is onset , there is some sample dependence and some differences between powders and single crystals . as noted by ye _
et al_. , @xcite powder measurements of pmn yield a different slope for the thermal expansion measurements than do those of dkhil _
et al_. @xcite also , studies using neutron strain scanning techniques found different thermal expansion coefficients as a function of depth in single crystal samples .
@xcite therefore , part of the discrepancy observed between different samples may be associated with surface effects . in this regard
, we also note the presence of a change in the slope of the strain curve between 400k and 500k ; since this change is not observed in other pmn studies , we believe this feature to be sample - dependent and thus extrinsic . the phase - shifted model of polar nanoregions proposed by hirota _
et al_. provides a plausible starting point from which to understand the anomalous invar - like behavior in pmn below the burns temperature @xmath9 .
this model is based on the observation that the pb , mg , nb , and o ionic displacements obtained from the diffuse scattering measurements of vakhrushev _
et al_. @xcite on pmn can be decomposed into a center - of - mass conserving component , consistent with the condensation of a soft , transverse optic mode , and a uniform scalar component , corresponding to an acoustic phase shift . in so doing hirota _
et al_. were able to reconcile the discrepancies between the structure factors of the diffuse scattering and the soft to modes and , in particular , to explain the weakness of the diffuse scattering intensity observed in the vicinity of the ( 200 ) bragg peak .
@xcite the idea that the pnr are uniformly displaced with respect to the underlying cubic lattice in a direction parallel to the pnr polarization has already been used by gehring _ et al_. to explain the persistence of the strong diffuse scattering in a single crystal of pzn-8%pt against large electric fields applied along [ 001 ] .
@xcite in that study it was shown that the diffuse scattering near ( 003 ) , which measures the projections of the ionic displacements along [ 001 ] , decreases under an external [ 001 ] field as expected .
by contrast , the diffuse scattering near ( 300 ) , which measures the projections of the ionic displacements along [ 100 ] , i. e. perpendicular to the field direction , remains unaffected even for field strengths up to 10kv / cm .
thus a uniform polar state is not achieved .
this surprising behavior can be understood if one assumes that the electric field applied along [ 001 ] removes or reduces the pnr shifts along [ 001 ] while preserving those along [ 100 ] .
the diffuse scattering should then decrease anisotropically with field , as is observed . in the context of the invar - like behavior of pmn
below @xmath9 , we speculate that such uniform shifts of the pnr could effectively stabilize the lattice against subsequent thermal expansion at lower temperature .
such a scenario is consistent with the fact that both pmn-10%pt and pmn-20%pt retain cubic structures down to low temperature when examined with high @xmath14-resolution neutron diffraction methods .
we have presented a comprehensive neutron study of the diffuse scattering in the relaxor pmn . we have greatly extended the timescales of previous neutron measurements by taking data on three different spectrometers , bt9 , spins , and the hfbs , which provide elastic energy resolutions of 500@xmath46ev , 120@xmath46ev , and 0.4@xmath46ev hwhm , respectively . while the backscattering data represent a 300-fold improvement in energy resolution over those obtained by hiraka _
et al_. @xcite on spins , both yield an onset temperature of @xmath10k for the diffuse scattering .
this indicates that the pnr in pmn are static on timescales of at least 2ns below 420k , but are apparently dynamic at higher temperatures .
thus the true burns temperature in pmn , which was originally interpreted by burns and dacol as the condensation temperature of static regions of local , randomly - oriented , nanometer - scale polarization , @xcite is @xmath10k , not 620k .
independent evidence of the existence of purely dynamic , short - range , polar correlations has been reported in pmn-55%pt , a composition with no strong diffuse scattering and thus no static pnr , by ko _ et al_. who observed phenomena that are typically related to the relaxation of dynamic pnr .
@xcite these include significant brillouin quasielastic scattering , a softening of the longitudinal acoustic phonon mode , and a deviation of the dielectric constant from curie - weiss behavior over an 80k temperature interval above @xmath21 .
previous measurements of the temperature dependence of the neutron diffuse scattering have been extended in this study to 900k , well above @xmath9 . in so doing
we have unambiguously established the existence of two types of diffuse scattering based on the observation of two markedly different temperature dependencies , one of which vanishes at @xmath9 and one of which does not .
we associate the strong , temperature - dependent , diffuse scattering with the formation of static , short - range , polar correlations ( pnr ) because of its well documented response to an external electric field , and because it first appears at the same temperature at which the soft ( polar ) to mode reaches a minimum in energy .
we associate the weak , temperature - independent , diffuse scattering , which shows no obvious change across either @xmath21 or @xmath9 , with chemical short - range order because it persists to extremely high temperature .
we further confirm the observations of hiraka _
et al_. , @xcite who first characterized the distinctly different reciprocal space geometries of both types of diffuse scattering , and we show that the bow - tie shape of the weak diffuse scattering persists up to 900k @xmath117 .
key effects of the strong , temperature - dependent , diffuse scattering , and thus of the pnr , on the low - energy lattice dynamics of pmn are also highlighted in this study .
the neutron inelastic measurements on pmn-60%pt by stock _
et al_. prove conclusively that pnr can not be the origin of the anomalous broadening of long - wavelength to modes observed in pmn , pzn , and other compounds , also known as the waterfall effect , because pmn-60%pt exhibits the same effect but no strong diffuse scattering ( no pnr ) .
@xcite by contrast , many studies have shown that pnr do broaden the ta modes in pmn @xcite and in pzn-@xmath20pt , @xcite but that such effects are absent in compositions with no pnr such as pmn-60%pt .
@xcite our cold neutron data show that these effects are @xmath14-dependent . whereas long - wavelength ta modes with reduced wave vectors @xmath31@xmath12 remain well - defined and exhibit a nearly constant energy width ( lifetime ) from 100k to 900k , shorter wavelength ta modes with reduced wave vectors @xmath32@xmath12 broaden substantially , with the maximum broadening occurring at @xmath118k .
this result motivates a very simple physical picture in which only those acoustic phonons having wavelengths comparable to the size of the pnr are significantly scattered by the pnr ; acoustic modes with wavelengths much larger than the pnr are largely unaffected because they simply average over the pnr .
models describing this effect have been discussed elsewhere .
@xcite in particular , very recent work by xu _
et al_. has revealed the presence of a phase instability in pzn-4.5%pt that is directly induced by such a pnr - ta phonon interaction .
this is shown to produce a pronounced softening and broadening of ta modes in those zones where the diffuse scattering is strong , and provides a natural explanation of the enormous piezoelectric coupling in relaxor materials .
@xcite in addition , we have performed neutron measurements of the thermal expansion with extremely high @xmath14- and @xmath25-resolution over a broad temperature range extending well below @xmath21 and far above @xmath119k . in agreement with many other studies ,
our single crystal samples of pmn exhibit little or no thermal expansion below @xmath9 , behavior that is reminiscent of the invar effect , but an unusually large thermal expansion coefficient of @xmath151/k above @xmath9 , where the strong diffuse scattering is absent .
the crossover between null and large coefficients of thermal expansion coincides closely with @xmath9 , which suggests that the appearance of static pnr strongly affects the thermal expansion in pmn and thus provides a structural signature of the formation of short - range , polar correlations in zero field .
the model of uniformly displaced , or phase - shifted , pnr proposed by hirota _
et al_. , which successfully predicts the anisotropic response of the strong diffuse scattering to an electric field , offers a simplistic , yet plausible , framework in which to understand this anomalous behavior .
finally , it is satisfying to note that the revised value of @xmath89k is consistent with the dielectric susceptibility data of viehland _
et al_. , from which a curie - weiss temperature @xmath120k was derived .
@xcite such good agreement between @xmath9 and @xmath121 solidifies our identification of the strong diffuse scattering with the condensation of the soft to mode , which reaches a minimum frequency at @xmath118k .
however this begs the question of how one should interpret the original burns temperature of @xmath122k . at present
there are two broadly divergent opinions on this issue , one of which considers @xmath122k to be a meaningful temperature scale in pmn , and one of which does not .
as it turns out , this debate is closely tied to another on how many temperature scales are needed to describe the physics of relaxors .
we offer no final resolution to this discussion .
instead , we close our paper with a brief summary of the primary studies supporting these contrasting points of view , which is by no means comprehensive , so that the readers may draw their own conclusions .
a number of experimental studies report evidence of either structural or dynamic changes in pmn in the temperature range 600k 650k , starting with the optical index of refraction measurements of burns and dacol .
@xcite siny and smirnova were the first to observe strong , first - order raman scattering in pmn at high temperatures , which , being forbidden in centrosymmetric crystals , implied the presence of some type of lattice distortion .
@xcite in 2002 vakhrushev and okuneva calculated the probability density function for the pb ion in pmn within the framework of the spherical layer model using x - ray powder diffraction data .
@xcite it was shown that this probability density evolves from a single gaussian function centered on the perovskite @xmath123-site to a double gaussian form between 635k and 573k , and that the positions of maximum density for the lead ion follow a power law @xmath124 with @xmath125k .
this picture was developed further by prosandeev _
et al_. within a model that ascribed the changes in the lead probability density function and subtle variations in the thermal expansion near 620k to a crossover from soft - mode to order - disorder dynamics .
@xcite in 2004 egami _
et al_. first proposed that at very high temperatures the vibrations of the oxygen octahedra are sufficiently faster than those of the pb ions that , on average , a high - symmetry , local environment is obtained , whereas at temperatures below @xmath126k the pb and o ions become dynamically correlated .
@xcite this led to the subsequent reinterpretation of the burns temperature ( @xmath127k ) as being the point below which dynamic pnr first form , which was based on the dynamic pdf measurements of dmowski _ et al_. obtained with an elastic energy resolution of 4mev at six temperatures ( 680k , 590k , 450k , 300k , 230k , and 35k ) .
@xcite interestingly , after integrating their data from -5mev to + 5mev , dmowski _
et al_. saw evidence of a third temperature scale in pmn of order 300k , and thus distinct from both @xmath128k and @xmath129k , which they associated with the formation of static pnr .
similar ideas have very recently been put forth by dkhil _
et al_. based on extremely interesting acoustic emission and thermal expansion measurements , and also in a general summary written by toulouse .
@xcite a different approach was taken by stock _
et al_. who proposed a unified description of the lead - based relaxors on the basis of striking similarities between pmn and the zn - analogue pzn , both of which display strong , temperature - dependent , diffuse scattering ; identical soft mode dynamics ; yet no long - range , structural distortion at low temperature in zero field .
@xcite arguing by analogy with magnetic systems , stock _
et al_. considered a three - dimensional heisenberg model with cubic anisotropy in the presence of random fields in which the heisenberg spin corresponds to the local ferroelectric polarization , the cubic anisotropy represents the preferential orientation of the polarization along the < 111 > axes , and the isotropic random magnetic field corresponds to the randomly - oriented local electric fields that originate from the varying charge of the @xmath64-site cation . following a suggestion by aharony , stock _
et al_. considered the case where the heisenberg term dominates over the random field term , and the cubic anisotropy term is the weakest of the three . in this picture there would be just two distinct , static , temperature scales . for @xmath130 ,
the cubic anisotropy is irrelevant and therefore the system should behave like a heisenberg system in a random field . in this case
the excitation spectrum is characterized by goldstone modes and therefore no long - range order is expected in the presence of random fields .
@xcite instead the system forms polar nanoregions in a paraelectric background .
the second temperature scale @xmath21 appears at low temperatures where the cubic anisotropy term becomes important , and in this limit the system should resemble an ising system in the presence of a random field .
this model thus explains the local ferroelectric distortion characterized by the recovery of the soft - optic mode and , although a 3d ising system in equilibrium should display long - range order in the presence of small random fields , as is observed in magnetic systems , nonequilibrium effects with long time scales become dominant .
the presence of such nonequilibrium effects may explain the lack of long - range ferroelectric order in pmn and pzn as well as the history dependence of physical properties such as the linear birefringence .
the phase - shifted nature of the polar nanoregions may also create another energy barrier against the ordered phase at @xmath21 .
et al_. are also able to explain the physics of compounds beyond the mpb , such as pmn-60%pt , within this model , for which only one temperature scale ( @xmath21 ) exists .
as the ti concentration ti increases , the relaxor phase diagram crosses over to a ferroelectric regime , and this can be understood as an increase in the strength of the cubic anisotropy term that is simultaneously accompanied by a reduction of the random fields as must occur in the limit of pure pbtio@xmath4 .
@xcite it should be mentioned that other studies have invoked random fields to explain the static and dynamic properties of relaxors .
@xcite independent of the validity of either of the two pictures summarized above , the seminal finding of our study of remains that the strong diffuse scattering in pmn first appears at a temperature that depends sensitively on the energy resolution of the measurement .
this fact inevitably raises interesting questions about the significance of the previous value of the burns temperature ( @xmath122k ) . if the strong diffuse scattering in pmn results from the soft to mode , then other techniques based on x - ray diffraction , thermal neutron scattering , or neutron pdf , which provide comparatively much coarser energy resolution , will be unable to discriminate between low - energy , dynamic , short - range polar correlations and truly static ( @xmath40 ) pnr because any low - energy , polar correlations will be integrated into the elastic channel by the broad energy resolution .
thus as the to mode softens , the associated low - energy , polar , spectral weight will fall within the energy resolution at temperatures above @xmath118k , and the net result will be an artificially higher value of @xmath9 that increases with the size of the energy resolution .
this effect should be especially pronounced for phonon modes that are broad in energy ( i. e. that have short lifetimes ) , and this is clearly the case for the soft to mode in pmn , which exhibits a nearly overdamped lineshape and a minimum frequency precisely at @xmath118k . while it is true that the structure factor of the strong diffuse scattering is inconsistent with that of the soft to mode , the phase - shifted model of hirota _
et al_. provides a way to reconcile this discrepancy . in this respect , we simply suggest that the previous value of @xmath131k might not represent a physically meaningful temperature scale in pmn .
future studies examining the values of @xmath9 in other relaxor compounds with improved energy resolution are thus of interest . our reassessment of the burns temperature @xmath9 immediately clarifies an intimate relationship between the formation of static , short - range polar correlations and the consequent effects on both the low - energy lattice dynamics and structure of pmn .
cold neutron triple - axis and backscattering spectroscopic methods conclusively show the existence of static , short - range polar correlations , only below @xmath89,k on timescales of at least 2ns .
thermal neutron measurements of the lattice dynamics reflect the presence of these static pnr through the presence of a distinct minimum in both the soft to and ta mode energies , both of which occur at 420k .
the effect of pnr on the lattice dynamics is evident only for ta modes having wave vectors of order 0.2@xmath12 , a fact that could be exploited to determine the size of the pnr . at the same time
an enormous change in the coefficient of thermal expansion is seen near @xmath9 , below which the crystal lattice exhibits invar - like behavior .
we would like to thank a. bokov , y. fujii , k. hirota , d. phelan , s. shapiro , s. wakimoto , and g. xu for stimulating discussions .
this study was supported in part by the u. s. - japan cooperative neutron - scattering program , the natural sciences and engineering research council ( nserc ) of canada , the national research council ( nrc ) of canada , the japanese ministry of monbu - kagaku - shou , rfbr grants 08 - 02 - 00908 and 06 - 02 - 90088nsf , the u. s. dept . of energy under contract
de - ac02 - 98ch10886 , the u. s. office of naval research under grant no .
n00014 - 06 - 1 - 0166 , and by the nsf under grant no . dmr-9986442 .
we also acknowledge the u. s. dept . of commerce ,
nist center for neutron research , for providing the neutron scattering facilities used in this study , some of which are supported in part by the national science foundation under agreement no .
dmr-0454672 .
key engineering materials _ vols . *
155 - 156 * , 81 ( 1998 ) .
s .- e . park and t. r. shrout , j. appl . phys .
* 82 * , 1804 ( 1997 ) .
a.a . bokov and z .- g .
ye , j. mat .
science * 41 * , 31 ( 2006 ) .
vugmeister , phys . rev .
b * 73 * , 174117 ( 2006 ) .
g. burns and f. h. dacol , solid state commun . *
48 * , 853 ( 1983 ) ; _ ibid _ , phys .
b * 28 * , 2527 ( 1983 ) .
p. bonneau , p. garnier , e. husson , and a. morell , mat .
bull , * 24 * , 201 ( 1989 ) .
n. de mathan , e. husson , g. calvarin , j.r .
gavarri , a.w .
hewat , and a. morell j. phys : condens matter * 3 * , 8159 ( 1991 ) .
j. zhao , a. e. glazounov , q. m. zhang , and b. toby , appl .
phys . lett . * 72 * , 1048 ( 1998 ) .
k. hirota , s. wakimoto , and d.e .
cox , j. phys .
. jpn . * 75 * , 111006 ( 2006 ) .
r. blinc , v. laguta , and b. zalar , phys .
* 91 * , 247601 ( 2003 ) . v.
v. shvartsman and a. l. kholkin , phys .
b * 69 * , 014102 ( 2004 ) .
t. egami , s. teslic , w. dmowski , p. k. davies , and i .- w . chen , j. kor .
. soc . * 32 * , s935 ( 1998 ) ; t. egami , w. dmowski , s. teslic , p. k. davies , i. w. chen , and h. chen , ferroelectrics * 206 * , 231 ( 1998 ) .
jeong , t. w. darling , j. k. lee , th .
proffen , r. h. heffner , j.s .
park , k.s .
hong , w. dmowski , and t. egami , phys .
* 94 * , 147602 ( 2005 ) .
a. naberezhnov , s. b. vakhrushev , b. dorner , and h. moudden , eur .
j. b * 11 * , 13 ( 1999 ) . t. y. koo , p. m. gehring , g. shirane , v. kiryukhin , s. g. lee , and s. w. cheong , phys .
b * 65 * , 144113 ( 2002 ) . s. vakhrushev , a. naberezhnov , s. k. sinha , y. p. feng , and t. egami , j. phys .
chem . solids * 57 * , 1517 ( 1996 ) .
h. you and q. m. zhang , phys . rev .
lett . * 79 * , 3950 ( 1997 ) .
g. xu , g. shirane , j. r. d. copley , and p. m. gehring , phys . rev .
b * 69 * , 064112 ( 2004 ) .
g. xu , z. zhong , y. bing , z .- g . ye , and g. shirane , nature materials * 4 * , 887 ( 2005 ) . m. matsuura , k. hirota , p. m. gehring , z .- g .
ye , w. chen , and g. shirane , phys .
b * 74 * , 144107 ( 2006 ) .
g. xu , z. zhong , h. hiraka , and g. shirane phys .
b * 70 * , 174109 ( 2004 )
. s. b. vakhrushev , a. a. naberezhnov , n. m. okuneva , and b. n. savenko , phys .
solid state * 40 * , 1728 ( 1998 ) . c. stock , g. xu , p. m. gehring , h. luo , x. zhao , h. cao , j. f. li , d. viehland , g. shirane , phys . rev .
b * 76 * , 064122 ( 2007 ) . p. m. gehring , k. ohwada , and g. shirane , phys .
b * 70 * , 014110 ( 2004 ) .
g. xu , p. m. gehring , and g. shirane , phys .
b * 72 * , 214106 ( 2005 ) .
g. xu , p. m. gehring , and g. shirane , phys .
b * 74 * , 104110 ( 2006 )
. s. wakimoto , c. stock , r. j. birgeneau , z .- g . ye , w. chen , w. j. l. buyers , p. m. gehring , and g. shirane , phys . rev .
b * 65 * , 172105 ( 2002 ) .
h. hiraka , s .- h .
m. gehring , g. xu , and g. shirane , phys .
b * 70 * , 184105 ( 2004 ) . c. stock , h. luo , d. viehland , j. f. li , i. swainson , r. j. birgeneau , and g. shirane , j. phys . soc .
jpn . , * 74 * , 3002 ( 2005 ) .
j. hlinka , s. kamba , j. petzelt , j. kulda , c. a. randall , and s. j. zhang , j. phys .
matter * 15 * , 4249 ( 2003 ) .
s. n. gvasaliya , s. g. lushnikov , and b. roessli , cryst .
reports * 49 * , 108 ( 2004 ) .
s. n. gvasaliya , s. g. lushnikov , and b. roessli , phys .
b * 69 * , 092105 ( 2004 )
. the study by hlinka _
et al_. ( ref . ) employed a much narrower energy resolution of 30@xmath46ev full - width at half - maximum ( fwhm ) than did the studies of gvasaliya _
et al_. and hiraka _ et al_. ( 200@xmath46ev fwhm ) ( ref .
s and ) . b. p. burton , e. cockayne , s. tinte , and u. v. waghmare , phase transitions * 79 * , 91 ( 2006 ) . c. stock , d. ellis , i.p .
swainson , guangyong xu , h. hiraka , z. zhong , h. luo , x. zhao , d. viehland , r. j. birgeneau , and g. shirane , phys .
b. , * 73 * , 064107 ( 2006 ) .
g. xu , p. m. gehring , v. j. ghosh , and g. shirane , acta .
a*60 * , 598 ( 2004 ) .
a. meyer , r. m. dimeo , p. m. gehring , and d. a. neumann , rev .
instrum . , * 74 * , 27 59 ( 2003 )
. w. chen and z .- g .
ye , unpublished ; z .- g . ye , p. tissot , and h. schmid , mater .
bull . * 25 * , 739 ( 1990 ) .
s. b. vakhrushev , private communication .
s. b. vakhrushev , a. a. naberezhnov , n. m. okuneva , and b. n. savenko , phys . solid state * 37 * , 1993 ( 1995 ) .
s. vakhrushev , a. ivanov , and j. kulda , phys .
phys . * 7 * , 2340 ( 2005 ) .
d. la - orauttapong , j. toulouse , z .- g . ye , w. chen , r. erwin , and j. l. robertson , phys . rev .
b * 67 * , 134110 ( 2003 ) .
these authors show schematic diffuse scattering intensity contours in fig . 4 of their paper ( no data are shown ) for pzn-4.5%pt and pzn-9%pt that are wrong because they do not possess the required crystal lattice symmetry
. t. r. welberry , d. j. goossens , and m. j. gutmann , phys .
b * 74 * , 224108 ( 2006 ) .
welberry , m.j gutmann , hyungje woo , d. j. goossens , guangyong xu , and c. stock , j. appl .
38 * , 639 ( 2005 ) .
s. wakimoto , g.a .
samara , r.k .
grubbs , e.l .
venturini , l.a .
boatner , g. xu , g. shirane , s .- h .
lee , phys .
b * 74 * , 054101 ( 2006 ) .
m. pasciak , m. wolcyrz , and a. pietraszko , phys .
b * 76 * , 014117 ( 2007 ) .
b.p burton and e. cockayne , phys .
b * 60 * , r12542 ( 1999 ) .
laguta , m.d .
glinchuk , s.n .
nokhrin , i.p .
bykov , r. blinc , a. gregorovic , and b. zalar , phys .
b * 67 * , 104106 ( 2003 ) . c. boulesteix , f. varnier , a. llebaria , and e. husson , j. solid state . chem . * 108 * , 141 ( 1994 ) .
o. svitelskiy , j. toulouse , g. yong , and z .- g . ye , phys . rev .
b * 68 * , 104107 ( 2003 ) .
k. hirota , z .- g . ye , s. wakimoto , p. m. gehring , and g. shirane , phys .
b * 65 * , 104105 ( 2002 ) .
r. comes , m. lambert , and a. guinier , acta crystallogr .
a * 26 * , 244 ( 1970 ) .
s. n. gvasaliya , b. roessli , r. a. cowley , p. huber , and s. g. lushnikov , j. phys .
condens . matter * 17 * , 4343 ( 2005 ) .
s. wakimoto , c. stock , z .- g . ye , w. chen , p. m. gehring , g. shirane , phys . rev .
b * 66 * , 224102 ( 2002 ) . h. sompolinsky and a. zippelius , phys . rev b * 25 * , 6860 ( 1982 ) . p.
m. gehring , s .- e .
park , and g. shirane , phys .
lett . * 84 * , 5216 ( 2000 ) .
m. gehring , s. b. vakhrushev , and g. shirane , in _ fundamental physics of ferroelectrics 2000 : aspen center for physics winter workshop _ , edited by r. e. cohen , aip conf .
535 ( aip , new york , 2000 ) , p. 314 .
m. gehring , s. wakimoto , z .- g . ye , and g. shirane , phys .
87 * , 277601 ( 2001 ) .
m. gehring , s .- e .
park , and g. shirane , phys .
b * 63 * , 224109 ( 2001 ) .
l. a. shebanov , p.
p. kapostins , and j. a. zvirgds , ferroelectrics * 56 * , 53 ( 1984 ) . h. arndt and g. schmidt , ferroelectrics * 79 * , 149 ( 1988 ) .
b. dkhil , j. m. kiat , g. calvarin , g. baldinozzi , s.b .
vakhrushev , and e. suard , phys .
b * 65 * , 024104 ( 2001 ) .
agrawal , a. halliyal , and j. belsick , mat .
* 23 * , 159 ( 1988 ) .
m. gehring , w. chen , z .- g .
ye , and g. shirane , j. phys . condens .
matter * 16 * , 7113 ( 2004 ) .
h. w. king , m. yildiz , s. h. ferguson , d. f. waechter , and s. e. prasad , ferroelectric letters * * , 55 ( 2004 ) .
p. bonneau , p. garnier , g. calvarin , e. husson , j. r. gavarri , a.w .
hewat , and a. morell , j. sol
. state chem .
* 91 * , ( 1991 ) .
g . ye , y. bing , j. gao , a. a. bokov , p. stephens , b. noheda , and g. shirane , phys . rev .
b * 67 * , 104104 ( 2003 ) .
k. h. conlon , h. luo , d. viehland , j.f .
li , t. whan , j.h .
fox , c. stock , and g. shirane , phys .
b * 70 * 172204 ( 2004 )
. g. xu , p. m. gehring , c. stock , k. conlon , phase transitions * 79 * , 135 ( 2006 ) .
g. xu , d. viehland , j. f. li , p. m. gehring , and g. shirane , phys .
b * 68 * , 212410 ( 2003 ) . j
ko , s. kojima , a. a. bokov , and z .- g .
ye , appl .
. lett . * 91 * , 252909 ( 2007 ) .
g. xu , j. wen , c. stock , and p. m. gehring , nature mat . * 7 * , 562 ( 2008 ) .
d. viehland , s. j. jang , l. e. cross , and m. wuttig , phys .
b * 46 * , 8003 ( 1992 ) . i. g. siny and t. a. smirnova , ferroelectrics * 90 * , 191 ( 1989 ) . s. b. vakhrushev and n. m. okuneva , in _ fundamental physics of ferroelectrics _ , edited by r. e. cohen , aip conf . proc .
626 ( aip , new york , 2002 ) ,
t. egami , w. dmowski , i .- k . jeong , r. h. heffner , j .- s .
park , k .- s .
hong , m. hehlen , and f. trouw , in _ fundamental physics of ferroelectrics _ edited by r. e. cohen and p. m. gehring , williamsburg conf .
( washington , dc , 2004 ) , p. 58 .
( http://people.gl.ciw.edu/cohen/meetings/ferro2004/ferro2004abstractbook.pdf ) w. dmowski , s. b. vakhrushev , i .-
jeong , m. p. hehlen , f. trouw , and t. egami , phys .
lett . * 100 * , 137602 ( 2008 ) .
m. roth , e. mojaev , e. dulkin , p. gemeiner , b. dkhil , phys .
98 * , 265701 ( 2007 ) .
j. toulouse , ferroelectrics * 369 * , 203 ( 2008 ) . c. stock , r. j. birgeneau , s. wakimoto , j. s. gardner , w. chen , z .- g . ye , and g. shirane , phys .
b * 69 * , 094104 ( 2004 ) .
g. xu , z. zhong , y. bing , z .- g . ye , c. stock , and g. shirane , phys . rev .
b * 67 * , 104102 ( 2003 ) .
a. aharony , private communication .
v. westphal , w. kleemann , and m. d. glinchuk , phys .
lett . * 68 * , 847 ( 1992 ) .
r. pirc and r. blinc , phys .
b * 60 * , 13470 ( 1999 ) .
r. fisch , phys .
b * 67 * , 094110 ( 2003 ) . | we have used neutron scattering techniques that probe time scales from @xmath0s to @xmath1s to characterize the diffuse scattering and low - energy lattice dynamics in single crystals of the relaxor pbmg@xmath2nb@xmath3o@xmath4 from 10k to 900k .
our study extends far below @xmath5k , where long - range ferroelectric correlations have been reported under field - cooled conditions , and well above the nominal burns temperature @xmath6k , where optical measurements suggest the development of short - range polar correlations known as polar nanoregions " ( pnr ) .
we observed two distinct types of diffuse scattering .
the first is weak , relatively temperature independent , persists to at least 900k , and forms bow - tie - shaped patterns in reciprocal space centered on @xmath7 bragg peaks .
we associate this primarily with chemical short - range order .
the second is strong , temperature dependent , and forms butterfly - shaped patterns centered on @xmath7 bragg peaks .
this diffuse scattering has been attributed to the pnr because it responds to an electric field and vanishes near @xmath6k when measured with thermal neutrons .
surprisingly , it vanishes at 420k when measured with cold neutrons , which provide @xmath8 times superior energy resolution .
that this onset temperature depends so strongly on the instrumental energy resolution indicates that the diffuse scattering has a quasielastic character and demands a reassessment of the burns temperature @xmath9 .
neutron backscattering measurements made with 300 times better energy resolution confirm the onset temperature of @xmath10k .
the energy width of the diffuse scattering is resolution limited , indicating that the pnr are static on timescales of at least 2ns between 420k and 10k .
transverse acoustic ( ta ) phonon lifetimes , which are known to decrease dramatically for wave vectors @xmath11 @xmath12 and @xmath13 , are temperature independent up to 900k for @xmath14 close to the zone center .
this motivates a physical picture in which sufficiently long - wavelength ta phonons average over the pnr ; only those ta phonons having wavelengths comparable to the size of the pnr are affected .
finally , the pmn lattice constant changes by less than 0.001 below 300k , but expands rapidly at a rate of @xmath151/k at high temperature .
these disparate regimes of low and high thermal expansion bracket the revised value of @xmath9 , which suggests the anomalous thermal expansion results from the condensation of static pnr . |
the first dust in the universe plays critical roles in the subsequent formation processes of stars and galaxies .
dust grains provide additional pathways for cooling of gas in metal - poor molecular clouds through their thermal emission and formation of h@xmath13 molecules on the surface ( e.g. , cazaux & spaans 2004 ) .
in particular , the presence of dust decreases the values of the critical metallicity to @xmath14@xmath15 @xmath8 ( omukai et al .
2005 ; schneider et al . 2006 ; tsuribe & omukai 2006 ) , where the transition of star formation mode from massive population iii stars to low - mass population ii stars occurs .
since absorption and thermal emission by dust grains strongly depend on their composition , size distribution , and amount , it is essential to clarify the properties of dust in the early epoch of the universe , in order to elucidate the evolutional history of stars and galaxies .
dust grains at redshift @xmath16 are considered to have been predominantly produced in supernovae ( sne ) .
theoretical studies have predicted that dust grains of 0.12 @xmath9 and 1060 @xmath9 are formed in the ejecta of primordial type ii sne ( sne ii , todini & ferrara 2001 ; nozawa et al .
2003 ) and pair - instability sne ( pisne , nozawa et al . 2003 ; schneider et al .
2004 ) , respectively .
however , the newly formed dust is reprocessed via sputtering in the hot gas swept up by the reverse and forward shocks that are generated from the interaction between the sn ejecta and the surrounding medium .
thus , the size and mass of the dust can be greatly modified before being injected into the interstellar medium ( ism , bianchi & schneider 2007 ; nozawa et al .
2007 ) .
in this proceedings we present the results of the calculations for the evolution of newly formed dust within population iii sn remnants ( snrs ) , based on the dust formation model by nozawa et al .
we investigate the transport of dust and its processing by sputtering in the shocked hot gas , and report the size and amount of dust injected from sne into the ism .
it is also shown that a part of the surviving dust grains are piled up in the dense sn shell formed behind the forward shock and can enrich the gas in the dense shell up to 10@xmath610@xmath7 @xmath8 .
we suppose that newly condensed dust in the sn ejecta has significant influences on the elemental abundances of population ii.5 stars , that is , the second - generation stars formed in the dense shell of population iii snrs .
we first describe the models of calculations for the evolution of dust within snrs in short .
the time evolution of a snr is numerically solved by assuming spherical symmetry .
we adopt the hydrodynamic models of population iii sne ii with the progenitor mass of @xmath17 = 13 , 20 , 25 , and 30 @xmath9 and explosion energy of @xmath18 ergs by umeda & nomoto ( 2002 ) for the initial condition of gas in the ejecta . for the ambient medium
, we consider the uniform medium with the hydrogen number density of @xmath19 0.1 , 1 , and 10 @xmath2 .
for the model of dust in the he core , we adopt the dust grains formed in the unmixed ejecta by nozawa et al .
treating dust grains as test particles , we calculate the destruction and dynamics of dust by taking account into the size distribution as well as the spatial mass distribution of each grain species .
sio@xmath20 , and fe grains within the snr for @xmath21 @xmath22 and @xmath23 @xmath2 and ( _ b _ ) the time evolutions of their radii relative to the initial ones .
the evolution of dust with @xmath24 0.01 , 0.1 , and 1 @xmath4 m is shown by the dotted , dashed , and solid lines , respectively . the thick solid lines in ( _ a _ ) indicate the positions of the forward shock , the reverse shocks , and the surface of the he core .
, width=316 ] the results of calculations are shown in figure 1 .
figure 1_a _ shows the trajectories of c , mg@xmath13sio@xmath20 , and fe grains within the snr for @xmath21 @xmath9 and @xmath23 @xmath2 , and figure 1_b _ shows the time evolutions of their radii relative to the initial ones .
the positions of the forward shock , the reverse shock , and the surface of the he core are depicted by the thick solid lines in figure 1_a_. initially , newly formed dust grains are expanding with the cool gas in the ejecta , and thus they undergo neither gas drag nor processing by sputtering .
however , once they intrude into the hot gas swept up by the reverse shock penetrating into the ejecta , they acquire the high velocities relative to the gas and are eroded by kinetic and/or thermal sputterings .
the evolutions of dust grains after colliding with the reverse shock heavily depend on their initial radii @xmath0 and compositions .
for example , c grains formed in the outermost region of the he core encounter the reverse shock at 3650 yr . since the deceleration by the gas drag is more efficient for smaller grains , c grains with @xmath25 @xmath4 m quickly slow down . these small grains are eventually trapped in the hot gas ( @xmath26 k ) generated from the passage of the reverse and forward shocks and are completely destroyed by thermal sputtering .
c grains of @xmath27 @xmath4 m reduce their sizes by sputtering but can not be completely destroyed .
these grains are finally captured in the dense sn shell formed behind the forward shock at @xmath282@xmath2910@xmath30 yr , where the gas temperature is too low ( @xmath31 @xmath32 k ) to erode the dust grains by thermal sputtering .
c grains with @xmath33 @xmath4 m , which are not efficiently decelerated by the gas drag , can pass through the forward shock front and are injected into the ambient medium without being destroyed significantly .
mg@xmath13sio@xmath20 grains , which are formed in the o - rich layer , collide with the reverse shock at about 6000 yr , but the dependence of their evolutions on the initial radius is the same as c grains . on the other hand , fe grains formed in the innermost region of the ejecta hit the reverse shock after 13000 yr , and its 0.1 @xmath4m - sized grains are injected into the ambient medium because of the high bulk density . as is shown above ,
the small dust grains formed in the sn ejecta are predominantly destroyed by sputtering within the snr .
thus , the size distribution of the surviving dust is dominated by larger grains , compared to that at its formation .
note that the evolution of dust within snrs and thus the resulting size distribution of dust do not depend on the progenitor mass considered here , because their explosion energies are the same and the time evolutions of the gas temperature and density within snrs are similar . on the other hand
, the ambient gas density strongly affects the evolution of dust in snrs . the higher density in the ambient medium results in the higher density of the shocked gas and causes the efficient erosion and deceleration of dust due to more frequent collisions with the hot gas .
therefore , the initial radius below which dust is completely destroyed increases with increasing the ambient gas density and is 0.01 , 0.05 , and 0.2 @xmath4 m for @xmath34 0.1 , 1 , and 10 @xmath2 , respectively . as a result ,
the total mass of the surviving dust is smaller for the higher ambient density and ranges from 0.01 to 0.8 @xmath9 for @xmath19 10 to 0.1 @xmath2 , depending on the size distribution of dust formed in each sn .
lcccccc & @xmath35 & [ fe / h ] & [ c / fe ] & [ o / fe ] & [ mg / fe ] & [ si / fe ] + + 13 & @xmath36 & @xmath37 & @xmath38 & @xmath39 & @xmath40 & @xmath41 + 20 & @xmath42 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath47 + 25 & @xmath48 & @xmath49 & @xmath50 & @xmath51 & @xmath52 & @xmath53 + 30 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58 & @xmath59 + + 13 & @xmath60 & @xmath61 & @xmath50 & @xmath62 & @xmath63 & @xmath64 + 20 & @xmath65 & @xmath66 & @xmath67 & @xmath68 & @xmath69 & @xmath70 + 25 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 + 30 & @xmath77 & @xmath78 & @xmath79 & @xmath80 & @xmath73 & @xmath81 + + 13 & @xmath82 & @xmath83 & @xmath84 & @xmath85 & @xmath86 & @xmath87 + 20 & @xmath88 & @xmath89 & @xmath90 & @xmath91 & @xmath92 & @xmath93 + 25 & @xmath94 & @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 + 30 & @xmath100 & @xmath78 & @xmath101 & @xmath102 & @xmath103 & @xmath104 +
in this section we discuss the influence of dust on the elemental composition of population ii.5 stars that are expected to form in the dense shell of population iii snrs ( mackey et al .
2003 ; salvaterra et al .
2004 ; machida et al .
as shown in the last section , the dust grains surviving the destruction but not injected into the ism are piled up in the dense sn shell after 10@xmath3010@xmath105 yr .
this implies that the elemental composition of these piled - up grains can play an important role in the elemental abundances of population ii.5 stars .
furthermore , the existence of dust in the shell may enable the formation of stars with solar mass scales through its thermal emission if the gas is enriched to the critical metallicities ( omukai et al .
2005 ; schneider et al .
2006 ; tsuribe & omukai 2006 ) .
thus , we calculate the metallicities and metal abundances in the dense shell based on the elemental composition of piled - up grains , and compare with the observed abundance patterns of low - mass hyper - metal - poor ( hmp ) and ultra - metal - poor ( ump ) stars .
table 1 summarizes the calculated metallicities and elemental abundances in the dense shell of sn ii remnants as a function of the ambient gas density .
it should be noted that metallicity in the shell ranges from 10@xmath6 to 10@xmath7 @xmath8 , which is considered to cause the formation of stars with 0.11 @xmath9 .
in addition , most of calculated [ fe / h ] are in the range from @xmath10 to @xmath11 , which is in good agreement with those for hmp and ump stars
. we also plot in figure 2a the abundances of c , o , mg , and si relative to fe in the shell for @xmath106 and 1 @xmath2 .
we can see that the calculated abundances of mg and si showing 1100 times overabundances are consistent with the observations of hmp and ump stars .
because the elemental composition of dust piled up in the shell can reproduce the abundance patterns of fe , mg , and si in hmp and ump stars , it is considered that the transport of dust separated from metal - rich gas within snrs can be attributed to the elemental compositions of hmp and ump stars , if they are population ii.5 stars . and 1 @xmath2 ( _ filled triangles _ ) ; ( _ a _ ) derived from the elemental composition of the grains piled up in the shell , and ( _ b _ ) derived from the elemental composition of the piled - up grains and the gas outside the innermost fe layer .
for observational data of hmp and ump stars , the 3-d corrected abundances are adopted and are denoted by plus ( he0107 - 5240 with [ fe / h ] = -5.62 , collet et al .
2006 ) , asterisk ( he1327 - 2326 with [ fe / h ] = -5.96 , frebel et al .
2008 ) , and cross ( he0557 - 4840 with [ fe / h ] = -4.75 , norris et al .
2007 ) . , title="fig:",width=249 ] and 1 @xmath2 ( _ filled triangles _ ) ; ( _ a _ ) derived from the elemental composition of the grains piled up in the shell , and ( _ b _ ) derived from the elemental composition of the piled - up grains and the gas outside the innermost fe layer . for observational data of hmp and ump stars , the 3-d corrected abundances are adopted and are denoted by plus ( he0107 - 5240 with [ fe / h ] = -5.62 , collet et al .
2006 ) , asterisk ( he1327 - 2326 with [ fe / h ] = -5.96 , frebel et al .
2008 ) , and cross ( he0557 - 4840 with [ fe / h ] = -4.75 , norris et al .
2007 ) . , title="fig:",width=249 ] however , as can be seen from figure 2@xmath107 , no model considered here can reproduce @xmath108@xmath109 times excesses of c and o observed in hmp stars .
one of the reasons is that in the calculation we assumed that the metal - rich gas in the sn ejecta does not mix with the gas in the shell
. then we examine the abundance patterns in the shell by assuming that besides the piled - up grains , the gas outside the innermost fe layer in the ejecta is incorporated into the shell .
the results are shown in figure 2_b_. in this case we can reproduce the very large overabundances of c and o , but the excesses of mg and si are too large ( @xmath110100 times ) to agree with the observations . however , it could be possible to reproduce the abundance patterns of refractory elements observed in hmp stars unless the si - mg - rich layer is mixed into the shell .
unfortunately , it has been still debated what extent of the gas in the ejecta can mix into the sn shell when population ii.5 stars form .
nevertheless , we can conclude that newly formed dust in a population iii sn can have great impacts on the stellar mass and metal abundance of population ii.5 stars , if the metal - rich gas is not significantly incorporated into the dense gas shell . +
this work has been supported in part by world premier international research center initiative ( wpi initiative ) , mext , japan and by the grant - in - aid for scientific research of the japan society for the promotion of science ( 18104003 , 19740094 ) . | we investigate the evolution of dust formed in population iii supernovae ( sne ) by considering its transport and processing by sputtering within the sn remnants ( snrs ) .
we find that the fates of dust grains within snrs heavily depend on their initial radii @xmath0 . for type ii snrs expanding into the ambient medium with density of @xmath1 @xmath2 ,
grains of @xmath3 @xmath4 m are detained in the shocked hot gas and are completely destroyed , while grains of @xmath5 @xmath4 m are injected into the surrounding medium without being destroyed significantly .
grains with @xmath0 = 0.050.2 @xmath4 m are finally trapped in the dense shell behind the forward shock .
we show that the grains piled up in the dense shell enrich the gas up to 10@xmath610@xmath7 @xmath8 , high enough to form low - mass stars with 0.11 @xmath9 .
in addition , [ fe / h ] in the dense shell ranges from @xmath10 to @xmath11 , which is in good agreement with the ultra - metal - poor stars with [ fe / h ] @xmath12 .
we suggest that newly formed dust in a population iii sn can have great impacts on the stellar mass and elemental composition of population ii.5 stars formed in the shell of the snr . |
ultracold atoms in optical lattices offer unprecedented possibilities of controlling quantum matter and mimicking the systems of condensed - matter and high - energy physics @xcite . particularly fascinating
is the possibility to study ultracold atoms under the influence of strong artificial abelian and non - abelian magnetic " fields . the experimental realization of artificial abelian magnetic " fields , which reproduce the physics of electrons in strong magnetic fields , is currently achieved through diverse schemes : for atoms in a trap the simplest way is to rotate the trap @xcite , while for atoms in optical lattices
this can be accomplished by combining laser - assisted tunneling and lattice acceleration methods @xcite , by the means of lattice rotations @xcite , or by the immersion of atoms in a lattice within a rotating bose - einstein condensate ( bec ) @xcite .
several phenomena were predicted to occur in these arrangements such as the hofstadter
butterfly " @xcite and the escher staircase " @xcite in single - particle spectra , vortex formation @xcite , quantum hall effects @xcite , as well as other quantum correlated liquids @xcite . as shown by one of us in ref .
@xcite , it is simple to generalize the scheme of jaksch and zoller for generating artificial abelian magnetic " fields @xcite in order to mimic artificial non - abelian magnetic " fields .
to this aim we have to consider atoms with more internal states ( flavors " ) .
the gauge potentials that can be realized using standard manipulations , such as laser - assisted tunneling and lattice acceleration , can have practically arbitrary matrix form in the space of flavors " .
in such non - abelian potentials , the single - particle spectrum generally depicts a complex structure termed by one of us hofstadter
moth " @xcite , which is characterized by numerous extremely small gaps .
the model of ref .
@xcite has stimulated further investigations , including studies of nontrivial quantum transport properties @xcite , as well as studies of the integer quantum hall effect ( iqhe ) for cold atoms @xcite , spatial patterns in optical lattices @xcite , modifications of the landau levels @xcite , and quantum atom optics @xcite .
one should note , however , that the @xmath0 gauge potentials proposed in ref .
@xcite and used in most of the following works are characterized by _ non - constant _ wilson loops : atoms performing a loop around a plaquette undergo a unitary transformation which depends linearly on one of the spatial coordinates .
although such gauge potentials are interesting _ per se _ , the features characterizing the hofstadter moth " result from this _ linear spatial dependence _ of the wilson loop , rather than from their non - abelian nature . indeed , the hofstadter moth "-
like spectrum may actually be found in the standard abelian case with a wilson loop proportional to @xmath1 ( see fig .
1 ) . two of us have shown that cold fermionic atoms trapped in optical lattices and subjected to artificial magnetic " fields should exhibit an iqhe @xcite . if a static force is applied to atoms , for instance by accelerating the lattice ,
the transverse hall conductivity gives the relation between this external forcing and the transverse atomic current through the lattice .
it has been shown that this transverse conductivity is quantized , @xmath2 , where @xmath3 is an integer and @xmath4 is planck s constant .
note that this quantity can be easily measured from density profiles , as shown recently by umucalilar _
. _ @xcite .
the quantization of @xmath5 occurs , however , only if the fermi energy of the system is located inside a gap of the single - particle spectrum . while the observation of the iqhe seems to be experimentally feasible in abelian
magnetic " fields , it is hardly so in the deeply non - abelian regime in which the gaps of the moth " become very small @xcite .
the question therefore arises whether the consideration of non - abelian gauge potentials characterized by a _
constant _ wilson loop could stabilize the spectral gaps and guarantee the robustness of the iqhe in ultracold fermionic gases and whether an anomalous iqhe , as observed in graphene , can exist in such systems . in this work we provide affirmative answers to these questions by considering the iqhe in a system that features a non - abelian gauge potential characterized by specific non - commutating constant components and by a _
constant _ wilson loop .
we calculate the energy spectrum and we obtain a robust band structure with well developed gaps , which differs drastically from the case of the gauge potential of ref . @xcite . in particular , we note the existence of van hove singularities in the density of states and we obtain their analytical expression .
we then evaluate the conductivity @xmath6 for neutral currents using topological methods : we express @xmath6 in terms of the topologically invariant chern numbers associated to each energy band @xcite .
we eventually present a salient fractal phase diagram which represents the integer values of the transverse conductivity inside the infinitely many gaps of the spectrum . in this way
, we show that the iqhe survives in the non - abelian regime , but undergoes strong modifications with striking similarity to the anomalous iqhe in graphene @xcite : the transverse conductivity suddenly changes sign due to the presence of van hove singularities and is , under certain conditions , anomalous because of conical energy spectra .
we consider a system of non - interacting two - component fermionic atoms trapped in a 2d optical square lattice of unit length @xmath7 , with sites at @xmath8 , with @xmath9 integers .
the non - interacting limit can be reached using feshbach resonances , or simply at low densities .
the optical potential is strong , so that the tight - binding approximation holds .
the schrdinger equation for a single particle subjected to an artificial gauge potential then reads @xmath10 where @xmath11 ( resp .
@xmath12 ) is the tunneling operator and @xmath13 ( resp .
@xmath14 ) is the tunneling amplitude in the @xmath1 ( resp .
@xmath15 ) direction . in the following ,
we use @xmath7 as the length , and @xmath16 as the energy units , and set @xmath17 , except otherwise stated .
the tunneling operators are related to the gauge potential according to @xmath18 .
here we consider a general non - abelian gauge potential @xmath19 where @xmath20 and @xmath21 are parameters , @xmath22 , @xmath23 are pauli matrices and @xmath24 is the number of abelian magnetic flux quanta per unit cell . in order to realize such a potential we may consider the method of ref .
however , the specific form of this gauge potential allows us to consider an even more practical scheme based on a generalization of the method currently developed by klein and jaksch @xcite .
we may use @xmath25k atoms in @xmath26 or @xmath27 hyperfine manifolds , or @xmath28li with @xmath29 . for
@xmath25k one should optically pump and restrict the atomic dynamics to the two lowest zeeman sublevels in each of the hyperfine manifolds .
one can then employ different lattice tiltings in the @xmath1 and @xmath15 directions to perform laser ( raman assisted ) tunnelings that change the internal states of the atoms ; this allows to control the parameters @xmath20 and @xmath21 , and fixes the tunneling rate .
finally , the immersion of the system in a rotating bec will allow to control @xmath24 @xcite . in experiments ,
one routinely reaches the values of ( laser assisted , or direct ) tunneling rates in the range of 5 - 10 khz ( @xmath30 0.5 @xmath31k ) , fermi temperatures of the same order , and temperatures @xmath32 0.2 @xmath33 50 - 100 nk ( see for instance @xcite ) .
+ the tunneling operators are @xmath34 unitary matrices , @xmath35 which act on the two - component wave function @xmath36 .
the single - particle hamiltonian is invariant under translations defined by the operators @xmath37 and @xmath38 under the condition that @xmath39 , where @xmath40 and @xmath41 are integers .
consequently , the system is restricted to a @xmath42 super - cell and one can express the wave function as @xmath43 , with @xmath44 a @xmath41-periodic function .
the wave vector @xmath45 belongs to the first brillouin zone , a @xmath46-torus defined as @xmath47 $ ] and @xmath48 $ ] .
the schrdinger equation then reduces to a generalized harper equation @xmath49
artificial gauge potentials generally induce the following non - trivial unitary transformation for atoms hopping around a plaquette of the lattice : u = u_x u_y ( m+1 ) u_x^ u_y^ ( m ) . in the presence of the gauge potential eq .
, atoms performing a loop around a plaquette undergo the unitary transformation : u= e^i 2 ^2 + 2 ^2 + 2 2 & 2 ^ 2 - i ^2 2 + -2 ^ 2 - i ^2 2 & ^2 + 2 ^2 - 22 .
if one sets @xmath50 or @xmath51 , where @xmath52 is an integer , the matrix @xmath53 is proportional to the identity and the system behaves similarly to the hofstadter model @xcite . when @xmath54 , where @xmath55 , one finds that @xmath56 and the system is equivalent to the @xmath57-flux model in which half a flux quanta is added in each plaquette @xcite . in these particular cases where @xmath58 , the system is in the abelian regime . for any other values of the parameters
@xmath20 and @xmath21 , the matrix @xmath59 is a non - trivial @xmath0 matrix and the system is non - abelian . + a gauge invariant quantity which characterizes the system is given by the wilson loop
@xmath60 it is straightforward to verify that the system is non - abelian when @xmath61 , and that @xmath62 . in fig .
[ wilson ] , where we show the wilson loop s magnitude as a function of the parameters , @xmath63 , we can easily identify the regions corresponding to the abelian ( @xmath64 ) and to the non - abelian regimes ( @xmath61 ) .
we note that the abelian @xmath57-flux regime is reached at a singular point , @xmath54 , where @xmath55 .
we also point out that the statement according to which the non - abelian regime is reached when @xmath65 \ne 0 $ ] , and which can be found in previous works @xcite , is incorrect : for the situation where @xmath54 , one finds that @xmath65 = 2 i \ ,
e^{2 i m \pi \phi } \sigma_z$ ] , while the system is abelian because of its trivial wilson loop , @xmath66 .
contrary to the non - abelian systems considered in previous works @xcite , we emphasize that the gauge potential eq
. leads to a wilson loop which does not depend on the spatial coordinates . in the following section ,
we show that this feature leads to energy spectra and fractal structures which significantly differ from the hofstadter moth " @xcite .
the energy spectrum can be obtained through direct diagonalization of eq . .
in the abelian regime corresponding to @xmath67 or @xmath68 , where @xmath55 , one finds @xmath41 doubly - degenerated bands for @xmath69 . in this particular case , the representation of the spectrum as a function of the flux @xmath24 leads to the fractal hofstadter butterfly " @xcite . for the other abelian case @xmath70
, the system behaves according to the @xmath57-flux lattice : the spectrum @xmath71 depicts a hofstadter butterfly " which is contained between @xmath72 $ ] , i. e. shifted by @xmath73 with respect to the original butterfly " , and the system remarkably describes zero - mass dirac particles @xcite . in the non - abelian regime , which is reached for arbitrary values of the parameters @xmath74 , the spectrum
is constituted of @xmath75 separated bands as illustrated in fig .
[ bandfig ] . for these general situations , the representation of the spectrum as a function of the flux
@xmath24 leads to new interesting features . as in the abelian case , one observes repetitions of similar structures at various scales . however , new patterns arise in the non - abelian case , as illustrated in fig .
[ flux ] for @xmath76 and in fig .
[ one ] for @xmath77 and @xmath78 .
it is worth noticing that for arbitrary values of the parameters @xmath74 , the spectra show well - developed gaps contrasting with the hofstadter moth " which appears in the non - abelian system proposed in ref .
we further notice that the spectrum is periodic with period @xmath79 and is symmetric with respect to @xmath80 and @xmath73 . in the non - abelian regime close to @xmath81
, one observes that conical intersections are preserved in the energy spectrum .
as shown in the next section , the particles behave similarly to dirac particles in this non - abelian region and the system exhibits an anomalous quantum hall effect .
we eventually note that when the flux @xmath82 , the density of states reveals several van hove singularities at the energies @xmath83 and @xmath84 , where @xmath85 . as the flux increases , these singularities evolve and generally merge .
we evaluate the linear response of the system described by eq . to an external force ( lattice acceleration ) applied along the @xmath15 direction and we evaluate the transverse conductivity @xmath86 using kubo s formula . following the method of ref .
@xcite , one can generalize the well - known tknn expression @xcite to the present non - abelian framework , yielding @xmath87 where @xmath88 is the @xmath89th component of the wave function corresponding to the band @xmath90 such that @xmath91 , and @xmath92 refers to the first brillouin zone of the system .
the fermi energy @xmath93 is supposed to lie within a gap of the spectrum .
the transverse conductivity is then given by the contribution of all the states filling the bands @xmath94 situated below this gap .
. conceals a profound topological interpretation for the transverse conductivity based on the fibre bundle theory @xcite . in the present framework ,
such bundles are conceived as the product of the parameter space @xmath92 with the non - abelian gauge group @xmath0 .
this product space , which is supposed to be locally trivial but is generally expected to twist globally , is characterized by the non - abelian berry s curvature =( _
k_x ^y-_k_y ^x + [ ^x , ^y ] ) dk_x dk_y , where @xmath95 is the berry s connection
. the triviality of the fibre bundle is measured by the chern number c(e_)= _ ^2 tr , which is a topological invariant and is necessarily an integer .
note that each band @xmath90 is associated to a specific fibre bundle , on which a chern number is defined .
one eventually finds that the hall - like conductivity eq .
is given by a sum of integer chern numbers , _
x y= - _ e _ < e _ f c(e _ ) .
[ final ] as a consequence , the transverse hall - like conductivity of the system evolves by steps corresponding to integer multiples of the inverse of planck s constant and is robust against small perturbations .
the evaluation of these topological invariants leads to a complete understanding of the iqhe which takes place in the present context .
the aim is then to compute the chern number associated to each band @xmath96 of the spectrum .
this computation can be achieved numerically thanks to an efficient method developed by fukui _
_ @xcite and which can be applied to our specific system .
this method is summarized as follows : the brillouin zone @xmath92 , defined by @xmath47 $ ] and @xmath97 $ ] , is discretized into a lattice constituted by points denoted @xmath98 . on the lattice
one defines a curvature @xmath99 expressed as _ 12 ( _ l)= u_1 ( _ l ) u_2 ( _ l + ) u_1 ( _
l + ) ^-1 u_2 ( _ l)^-1 , where the principal branch of the logarithm with @xmath100 is taken , @xmath101 is a unit vector in the direction @xmath31 , and u _ ( _ l)= _ j u_j ( _ l ) u_j ( _ l+ ) / _ ( _ l ) , defines a link variable with a normalization factor @xmath102 such that @xmath103 .
the chern number associated to the band @xmath90 is then defined by c= _
l _ 12 ( _ l ) .
this method ensures the integral character of the chern numbers and holds for non - overlapping bands . in the situations where the spectrum reveals band crossings ,
a more general definition of the link variables @xmath104 has been proposed in ref .
@xcite .
we first compute the chern numbers for a specific case , illustrated in fig .
[ bandfig ] . for @xmath105 and @xmath106 ,
the chern numbers associated to the six bands are respectively @xmath107 . according to eq . , the transverse conductivity s values associated to the 5 gaps are @xmath108 as shown in fig .
[ bandfig ] .
the phase diagram describing the iqhe for our model can eventually be drawn . in this diagram
we represent the quantized transverse conductivity as a function of the fermi energy @xmath93 and flux @xmath24 . here
we illustrate a representative example of such a phase diagram which was obtained for @xmath109 , @xmath110 ( cf .
[ one ] ) .
this striking figure differs radically from the phase diagrams obtained by osadchy and avron in the abelian case @xcite since the chern numbers associated to the gaps are no longer satisfying a simple diophantine equation @xcite .
consequently , the measurement of the transverse conductivity in this system should show a specific sequence of robust plateaus , heralding a new type of quantum hall effect .
this new effect is comparable to the iqhe observed in si - mosfet or the anomalous iqhe observed in graphene in the low flux " regime @xmath111 corresponding to experimentally available magnetic fields . in this regime ,
the quantized conductivity evolves monotonically by steps of one between sudden changes of sign across the aforementioned van hove singularities ( see fig .
[ one ] ) .
moreover , in the vicinity of @xmath112 , the quantized conductivity increases by double integers because of dirac points in the energy spectrum , in close similarity with the anomalous iqhe observed in graphene .
summarizing , we have proposed in this paper how to realize in cold atomic systems a textbook example of non - abelian gauge potential characterized by a _ constant _ wilson loop .
our main result is that despite the coupling between the different flavor " components of the single - particle wave functions , the spectrum exhibits well - developed gaps of order of 0.1 - 0.2@xmath113 , i.e. about 50 - 100 nk .
the iqhe survives in the deeply non - abelian regime and acquires a unique character specific to the non - abelian nature of the gauge fields .
it is characterized by a particular sequence of robust plateaus corresponding to the quantized values of the transverse conductivity . moreover
, the non - abelian coupling induces controllable van hove singularities as well as an anomalous hall effect , similar to the effect induced by the hexagonal geometry in graphene .
experimental observation of this distinctive effect requires to achieve @xmath114 smaller than the gaps , i.e. of order of 10 - 50 nk , which is demanding but not impossible .
the main experimental challenge consists here in combining several established methods into one experiment : laser assisted tunneling @xcite , bec immersion @xcite , and density profile measurements @xcite .
we acknowledge support of the eu ip programme scala , esf - mec euroquam project fermix , spanish mec grants ( fis 2005 - 04627 , conslider ingenio 2010 qoit ) , the belgian federal government ( iap project
nosy `` ) , and the ' ' communaut franaise de belgique `` ( contract ' ' arc `` no
. 04/09 - 312 ) and the f.r.s .- fnrs belgium .
m.l acknowledges erc adg ' ' quagatua " .
n. g. thanks s. goldman for his comments , pierre de buyl for its support and icfo for its hospitality .
a.k . acknowledges support of the polish government research grant for 2006 - 2009 and thanks j. korbicz and c. menotti for discussions .
the authors thank d. jaksch , j. v. porto , and c. salomon for their valuable insights in various aspects of this work . | nowadays it is experimentally feasible to create artificial , and in particular , non - abelian gauge potentials for ultracold atoms trapped in optical lattices . motivated by this fact
, we investigate the fundamental properties of an ultracold fermi gas in a non - abelian @xmath0 gauge potential characterized by a _ constant _ wilson loop . under this specific condition ,
the energy spectrum exhibits a robust band structure with large gaps and reveals a new fractal figure .
the transverse conductivity is related to topological invariants and is shown to be quantized when the fermi energy lies inside a gap of the spectrum .
we demonstrate that the analogue of the integer quantum hall effect for neutral atoms survives the non - abelian coupling and leads to a striking fractal phase diagram .
moreover , this coupling induces an anomalous hall effect as observed in graphene . |
according to recent estimates , 8090% of the matter in the universe is non - baryonic . with the presence of extended dark matter halos in galaxies of all morphological types
now well - established , attention is focusing on comparing the detailed mass distribution with theoretical predictions of galactic dark halo structure .
the shape of the dark matter distribution is determined by the initial density perturbation spectrum , the coupled dynamical evolution of baryonic and non - baryonic constituents , and the nature of the dark matter itself .
therefore , its measurement represents a powerful diagnostic of fundamental astrophysical processes and parameters . the standard cold dark matter
( cdm ) model is highly successful in explaining the distribution of mass in the universe on scales ranging from galaxies on up , but is undergoing a critical re - examination due in large part to its confrontation with measurements of _ late - type _ galaxy mass distributions indicating that dark matter is less concentrated than expected . in this work @xcite
, we use a recent _ chandra _ x - ray observatory observation of ngc 4636 to constrain its dark matter distribution , with the goal of testing whether this `` concentration crisis '' applies to this ( _ early - type _ ) elliptical galaxy .
hot , extended distributions of x - ray emitting hot gas in hydrostatic equilibrium are crucial tracers of the gravitational potential in individual elliptical galaxies @xcite ; gravitational lensing is useful only in a statistical sense for a large sample of galaxies .
the physical properties of the hot gas can be accurately measured out to large radii where dark matter dominates and optical techniques are infeasible .
the effective resolution of the mass estimate essentially corresponds to that of the gas temperature profile ( see below ) . the unprecedented
_ chandra _ angular resolution enables one to do x - ray imaging spectroscopy of the hot ism on scales comparable to optical studies of the stars for the first time , thus yielding the relative distributions of luminous and dark matter . given a mass distribution , and the hot gas density distribution obtained from the measured x - ray surface brightness distribution , we solve the equation of hydrostatic equilibrium to derive the corresponding gas temperature profile .
the total mass model consists of three components an @xmath4 m@xmath2 central supermassive black hole @xcite , stars distributed as measured from _ hubble space telescope _ and ground - based photometry , and a dark matter density distribution parameterized by an asymptotic slope at the origin ( @xmath5 ) , a scale radius characterizing the transition to an @xmath6 decline ( as determined by the results of numerical simulations ) at large radius , and a normalization . for a given @xmath5 , the dark matter scale and normalization , and the stellar mass - to - light ratio , are varied to obtain a match with the composite temperature profile measured by _
chandra _ and the _ xmm - newton _ observatory ( figure 1 ) .
dark matter density cusp ; solid curve ) , and best - fit model with dark matter density core ( broken curve ) . ]
models with constant mass - to - light ratios and those where the dark matter density inner slope , @xmath5 , is as steep as the singular isothermal sphere value of @xmath7 are clearly ruled out .
we derive accurate constraints on the total mass distribution from 0.735 kpc that are robust to the assumed value of @xmath5 .
the total mass increases as @xmath0 ( more quickly than the stars ) to a good approximation over this range in radii , attaining a total of @xmath8 m@xmath2 ( corresponding to @xmath3 in solar units ) at the outermost point we consider ( figure 2 ) .
there are degeneracies with respect to the relative distributions of dark and luminous matter .
we find acceptable models with dark matter cusps ( e.g. , @xmath9 ) and with cores ( @xmath10 ) .
the former imply lower stellar mass to light ratios ( the allowed range : @xmath11 , consistent with studies of the stellar population ) and , in some cases , the predominance of dark matter all the way into the nucleus of the galaxy .
the dark matter fraction within the half - light radius ( @xmath12 kpc ) can be as high as 0.8 and `` central '' ( actually , the average over the inner 700 pc ) dark matter density as high as 4.0 m@xmath2 pc@xmath13 ; the cored models define lower limits to these quantities of 0.5 and 0.16 m@xmath2 pc@xmath13 , respectively .
even the lowest allowed central density is an order of magnitude greater than the previously estimated values for other types of galaxies ( figure 3 ) that precipitated the consideration of alternatives to cdm .
comparison of non - parametric measures of the dark matter concentration with expectations based on standard cdm numerical simulations indicate that dark matter in ngc 4636 is more compact than average for its mass scale , but within the expected scatter .
this is consistent with the findings of statistical studies of the gravitational lensing properties of elliptical galaxies , though it appears that ngc 4636 is unusually dark matter dominated . many of the proposed scenarios for reducing the dark matter concentration in late - type galaxies are at odds with our results .
it is instructive to consider the two classes of models illustrated in the above figures from this perspective those with flat dark matter cores , and those with cuspy dark matter cores ( that provide better fits to the x - ray data ) interpreted as contracting from initially flat cores due to baryonic infall . for the flat - core models , the high central dark matter mass density and large cores are contrary to the expected scaling relations for self - interacting dark matter ( sidm ) and other models where dark matter structure is driven by dark matter particle interaction . for the cuspy - core models the central phase - space density , conserved during adiabatic contraction , is too high .
alternatively , the velocity - dependent cross section can be fine - tuned so that interactions are ineffective at the higher mass ( and velocity ) scale of ellipticals ; although here too , one may run into phase space difficulties if ellipticals form from mergers of lower mass systems with interaction - induced dark matter cores .
scenarios where the mechanism for reducing the dark matter concentration becomes less effective with increasing mass scale are not ruled out , since ellipticals such as ngc 4636 represent the most massive galaxies with the deepest potential wells .
these include models that invoke warm dark matter , or the expansion of dark matter due to coupling with a powerful protogalactic starburst - driven baryonic outflow .
if this transitional mass is indeed on the giant elliptical galaxy scale , cuspy dark matter distributions in galaxy clusters are implied .
our result that the central dark matter density in ngc 4636 is @xmath14 the typical value for less massive galaxies contradicts the naive expectations of bottom - up hierarchical clustering where the most massive systems form latest and reflect the relatively low average density in the universe at that epoch .
a cold dark matter initial perturbation spectrum that is tilted or otherwise lacking in small - scale power may help explain ly-@xmath15 forest data @xcite , but likely exacerbates this contradiction
. recent work on the relation between the merging histories of galaxies and their morphology @xcite may be the key to resolving the dark matter concentration problem . a significant dispersion in dark matter concentration
is predicted by numerical simulations at any given mass range , but particularly on galaxy scales .
this results from the stochasticity of the merging process , and introduces the following bias .
galaxies of low central dark matter density represent relatively recently formed systems , and/or particularly fragile galaxies that experienced relatively tranquil assembly histories
these may correspond to disk galaxies where low dark matter central densities are indeed measured .
conversely , giant elliptical galaxies such as , or perhaps particularly , ngc 4636 may form at relatively high redshifts and undergo exceptionally prominent merger histories .
possible observational inconsistencies of the ( non - interacting ) cdm paradigm resulted in a renewed scrutiny and a proliferation of alternative models .
careful analysis of the observational situation in light of more detailed , and more deeply understood , theoretical models , as well as the shortcomings of the proposed alternatives seem to indicate that cdm is withstanding these recent challenges to remain the most viable model of structure in the universe ( see j. primack s contribution to this volume ) .
our results on the high concentration of dark matter in ngc 4636 evidently strengthens the case for cdm , while introducing other puzzles such as why ngc 4636 seems to be exceptionally dark matter dominated .
although one must exercise caution in generalizing from our results , the most stringent constraints on alternative models for dark halo structure , that presumably universally apply , emerge from studying such a , possibly extreme , system .
results , in progress , of similar investigations for other galaxies ( e.g. , ngc 1399 , ngc 4472 ) are anticipated with great interest .
9 loewenstein , m. and mushotzky , r. f. , astrophysical journal , submitted .
loewenstein , m. and white , r. e. , iii , astrophysical journal , 518 ( 1999 ) , 50 .
merritt , d. and ferrarese , l. , monthly notices of the royal astronomical society , 320 ( 2001 ) , l30 .
alam , s. m. k. , bullock , j. s. , and weinberg , d. h. , astrophysical journal , submitted .
wechsler , r. h. , bullock , j. s. , primack , j. r. , kravtsov , a. , and dekel , a. , astrophysical journal , 569 ( 2002 ) , 52 . | we determine the total enclosed mass profile from 0.7 to 35 kpc in the elliptical galaxy ngc 4636 based on the hot interstellar medium temperature profile measured using the _ chandra _ x - ray observatory , and other x - ray and optical data .
the total mass increases as @xmath0 to a good approximation over this range in radii , attaining a total of @xmath1 m@xmath2 ( corresponding to @xmath3 ) at 35 kpc .
we find that at least half , and as much as 80% , of the mass within the optical half - light radius is non - luminous , implying an exceptionally low baryon fraction in ngc 4636 .
the large inferred dark matter concentration and central dark matter density , consistent with the upper end of the range expected for standard cold dark matter halos , imply that mechanisms proposed to explain low dark matter densities in less massive galaxies are not effective in elliptical galaxies . |
over the last several years , there has been an enormous amount of interest in the physics of dilute fermi gases confined in magneto - optic traps@xcite . with the possibility of tuning the atomic scattering lengths from the repulsive regime to an attractive one using the feshbach resonance technique ,
there has been considerable experimental activity in looking for phenomenon such as superfluidity , and other phase transitions in these systems@xcite .
this has led to equally vigorous theoretical activity starting from the studies of so - called bec - bcs crossover physics@xcite , search for shell - structure in these systems@xcite , to the study of more complex phases@xcite .
as far as the spin of the fermions is concerned , most attention has been given to the cases of two - component gases which can be mapped to a system of spin-@xmath0 atoms@xcite .
therefore , in our opinion , a quantum - mechanical study of spin-@xmath0 fermions moving in a harmonic oscillator potential , and interacting via a pair - wise delta function potential , can help us achieve insights into the physics of dilute gases of trapped fermionic atoms . with the aforesaid
aims in mind , the purpose of this paper is to describe a fortran 90 computer program developed by us which can solve the hartree - fock equations for spin-@xmath0 fermions moving in a three - dimensional ( 3d ) harmonic oscillator potential , and interacting via delta - function potential .
a basis set approach has been utilized in the program , in which the single - particle orbitals are expanded as a linear combination of the 3d simple harmonic oscillator basis functions , expressed in terms of cartesian coordinates .
the program can solve both the restricted - hartree - fock ( rhf ) , and the unrestricted hartree - fock ( uhf ) equations , the latter being useful for fermi gases with imbalanced populations .
we would like to clarify , that as far as the applications of this approach to dilute fermi gases is concerned , at present it is not possible to reach the thermodynamic limit of very large @xmath4 , where @xmath4 is the total number of atoms in the trap .
however , we believe that by solving the hf equations for a few tens of atoms , one may be able to achieve insights into the microscopic aspects such as the nature of pairing in such systems . this program is an extension of an earlier program developed in our group , aimed at solving the time - independent gross - pitaevskii equation ( gpe ) for harmonically trapped bose gases@xcite .
thus the combined total program accompanying this paper can now solve for both bose and fermi systems , confined to move in a harmonic oscillator potential , with mutual interactions of the delta - function form . as with our earlier boson program , because of the use of a cartesian harmonic oscillator basis set , the new program can handle trap geometries ranging from spherical to completely anisotropic , and it can also account for those trap anharmonicities which can be expressed as polynomials in the cartesian coordinates . the nature of interparticle interactions , i.e. , whether they are attractive or repulsive , also imposes no restrictions on the program .
we note that yu _ _
et al.__@xcite _ _ have recently described a hartree - fock approach for dealing with two - component fermions confined in harmonic traps with spherical symmetry , employing a finite - difference - based numerical approach .
however , we would like to emphasize that , as mentioned earlier , our approach is more general in that it is not restricted to any particular trap symmetry .
apart from describing the program , we also present and discuss several of its applications . with the aim of exploring the shell - structure in trapped fermionic atoms , using our uhf approach we compute the addition energy for spherically trapped fermions for various particle numbers , and obtain results consistent with a shell - structure and hund s rule .
the remainder of the paper is organized as follows . in the next section
we discuss the basic theoretical aspects of our approach . in section [ sec - program ]
, we briefly describe the most important subroutines that comprise the new enlarged program
. section [ sec : install ] contains a brief note on how to install the program and prepare the input files . in section [ sec - results ]
we discuss results of several example runs of our program for different geometries . in the same section
, we also discuss issues related to the convergence of the procedure .
finally , in section [ sec - conclusions ] , we end this paper with a few concluding remarks .
we consider a system of @xmath4 identical spin-@xmath0 particles of mass @xmath5 , moving in a 3d potential with harmonic and anharmonic terms , interacting with each other via a pair - wise delta function potential .
the hamiltonian for such a system can be written as @xmath6 where @xmath7 represents the position vector of @xmath8th particle , @xmath9 represents the strength of the delta - function interaction , and @xmath10 denotes the one - particle terms of the hamiltonian@xmath11 where @xmath12 , @xmath13 and @xmath14 are the angular frequencies of the external harmonic potential in the @xmath15 , @xmath2 and @xmath3 directions , respectively , and @xmath16 represents any anharmonicity in the potential . in order to parametrize the strength of the delta - function interactions
, we use the formula @xmath17 in our program , where @xmath18 is the @xmath19-wave scattering length for the atoms .
next we will obtain the rhf and the uhf equations for the system . assuming that @xmath20 , and that the many - particle wave function of the system can be represented by a single closed - shell slater determinant , the rhf equations for the @xmath21 doubly occupied orbitals @xmath22 of the system are obtained to be@xcite@xmath23 similarly , for a system with @xmath24 up - spin ( @xmath25 ) fermions , and @xmath26 down - spin ( @xmath27 ) fermions @xmath28 , the uhf equations for the up - spin orbitals can be written as@xcite@xmath29 where @xmath30 and @xmath31 , represent the occupied orbitals corresponding to the up and the down spins , respectively .
similar to the the uhf equations for the down - spin orbitals can be deduced easily from eq .
( [ eq : uhf ] ) . as in our earlier work on the bosonic systems@xcite
, we adopt a basis - set approach and expand the hf orbitals in terms of the 3d harmonic oscillator basis functions .
this approach is fairly standard , and is well - known as the hartree - fock - roothan procedure in the quantum chemistry community@xcite .
thus , for the rhf case , the orbitals are expressed as@xmath32 where @xmath33 represents the coefficient corresponding to the @xmath34-th 3d harmonic oscillator basis function @xmath35 , in the expansion of the @xmath36-th occupied orbital @xmath37 , and @xmath38 is the total number of basis functions used . note that @xmath35 is itself a product of three linear harmonic oscillator eigenfunctions of quantum numbers @xmath39 @xmath40 , and @xmath41 .
therefore , a set of functions @xmath35 , for different values of @xmath39 @xmath40 , and @xmath41 , will constitute an orthonormal basis set , leading to an overlap matrix which is identity matrix .
for the uhf case , the corresponding expansion for up - spin particles is@xmath42 from which the expansion for the down - spin particles can be easily deduced . upon substituting eqs .
( [ eq : basis - rhf ] ) and ( [ eq : basis - uhf ] ) , in eqs .
( [ eq : rhf ] ) and ( [ eq : uhf ] ) , respectively , one can obtain the matrix forms of the rhf / uhf equations@xcite . as outlined in our earlier work@xcite , numerical implementation of the approach
is carried out in the so - called harmonic oscillator units , in which the unit of length is the quantity @xmath43 , and that of energy is @xmath44 .
the resulting matrix equation for the rhf case is @xmath45 where @xmath46 represents the column vector containing expansion coefficients @xmath47 of @xmath48 , @xmath49 is the corresponding energy eigenvalue , and the elements of the fock matrix @xmath50 are given by@xmath51 above@xmath52 expressed in terms of aspect ratios @xmath53 and @xmath54 , @xmath55 are the matrix elements of the anharmonic term in the confining potential , @xmath56 is a density - matrix element , and @xmath57 represents the 3d two - fermion repulsion matrix defined as@xmath58 each one of the @xmath59 matrices in eq .
( [ eq - jtilde ] ) , corresponding to the three cartesian directions , can be written in the form@xmath60 where @xmath61 is the corresponding cartesian coordinate in the harmonic oscillator units .
an analytical expression for @xmath62 can be found in our earlier work@xcite . in the uhf case ,
one obtains two matrix equations for the up / down - spin particles of the form@xmath63 where the @xmath64 represents the fock matrix for the up - spin particles given by @xmath65 , @xmath66 is the energy eigenvalue , and @xmath67 , are the elements of the down - spin density matrix . we can easily deduce the form of the fock equation for the down - spin particles from eq .
( [ eq - eigval - uhf ] ) . in our program ,
( [ eq - eigval - rhf ] ) and ( [ eq - eigval - uhf ] ) are solved employing the self - consistent field ( scf ) procedure , which requires the iterative diagonalization of the fock equations@xcite .
in this section we briefly describe the main program and various subroutines which constitute the entire module .
as mentioned in the introduction , the present program is an extension of our earlier program for bosons@xcite .
thus the new program , which compiles as _ trap.x _ , can solve for : ( a ) time - independent gross - pitaevskii equation for bosons , and ( b ) hartree - fock equations for fermions , confined in a trap .
therefore , most of the changes in the present program , as compared the earlier bosonic program , are related to its added fermionic hf capabilities .
however , we have also tried to optimize the earlier bosonic module of the program wherever possible .
a readme file associated with this program lists all its subroutines .
thus , in what follows , we will describe only those subroutines which are either new ( fermion related ) , or modified , as compared to the older bosonic code@xcite . for an account of the older subroutines not described here ,
we refer the reader to our earlier work@xcite . additionally , with the aim of making the calculations faster , in the present code
, we use the diagonalization routines of lapack library@xcite , which requires the linking of our code to that library .
therefore , for this program to work , the user must have the lapack / blas program libraries installed on his / her computer system .
the letter f or b has been included in parenthesis after the name of each subroutine to show whether the subroutine is useful for fermionic or bosonic calculations .
if it is applicable for both , we denote this by writing bf .
this is the main program of our package which reads the input data , dynamically allocates relevant arrays , and then calls other subroutines to perform tasks related to the remainder of the calculations .
the main modification in this program , as compared to its earlier version@xcite , is that it now allows for input related to fermionic hf calculations .
thus , the user now has to specify whether the particles considered are bosons or fermions .
if the particles considered are fermions , one has to further specify whether the rhf or the uhf calculations are desired . for the case of uhf calculations , the user also needs to specify the number of up- and down - spin orbitals . because of the dynamic array allocation throughout , no data as to the size of the arrays is needed from the user .
the program will stop only if it exhausts all the available memory on the computer .
there is one major departure in the storage philosophy in the present version of the code as compared to the previous one@xcite in that now only the lower / upper triangles of most of the real - symmetric matrices ( such as the fock matrix ) are stored in the linear arrays in the packed format .
this not only reduces the memory requirements roughly by a factor of two , but also leads to faster execution of the code .
this is the modified version of the old subroutine bec_drv , and is called from the main program oscl .
as its name suggests , it is the driver routine for performing : ( a ) calculations of the bose condensate wave function for bosons , or ( b ) solving the rhf / uhf equations for fermions .
apart from allocating a few arrays , the main task of this routine is to call either : ( a ) routines bose_scf or bose_steep depending upon whether the user wants to use the scf or the steepest - descent approach meant for solving the gpe@xcite , or ( b ) routines fermi_rhf or fermi_uhf depending on whether the rhf or uhf calculations are to be performed .
this subroutine solves the rhf equations for the fermions in a trap using the scf procedure , mentioned earlier .
its main tasks are as follows : 1 .
allocate various arrays needed for the scf calculations 2 .
setup the starting orbitals .
this is achieved by diagonalizing the one - particle part of the hamiltonian .
3 . perform the scf calculations . for this purpose , the two - particle integrals @xmath68 ( cf .
( [ eq - jtilde ] ) ) are calculated during each iteration@xcite .
if the user has opted for fock matrix / orbital mixing , it is implemented using the formula @xmath69 where @xmath70 is the quantity under consideration in the @xmath36-th iteration , and parameter @xmath71 quantifying the mixing is user specified .
thus , if fock matrix mixing has been opted , @xmath71 specifies the fraction of the new fock matrix in the total fock matrix in the @xmath36-th iteration .
if the user has opted for the orbital mixing , then each occupied orbital is mixed as per the formula above . the fock matrix constructed in each iteration
is diagonalized using the lapack routine dspevx@xcite , which can obtain a selected number of eigenvalues / eigenvectors of a real - symmetric matrix , as against traditional diagonalizers which calculate the entire spectrum of such matrices .
we use dspevx during the scf iterations to obtain only the occupied orbitals and their energies , thereby , leading to a much faster completion of the scf process in comparison to using a diagonalizer which computes all the eigenvalues / vectors of the fock matrix .
the occupied orbitals are identified according to the _ aufbau _ principle .
the total energy and the wave function obtained after every iteration are written in various data files so that the progress of the calculation can be monitored .
this process continues until the required precision ( user specified ) in the total hf energy is obtained . in structure and philosophy
this subroutine is similar to fermi_rhf , except that its purpose is to solve the uhf equations for interacting spin-@xmath0 fermions confined in a harmonic potential . because there are two separate fock equations corresponding to the up- and the down - spin fermions , the computational effort associated with this subroutine is roughly twice that of routine fermi_rhf .
this subroutine aims at solving the time - independent gpe for bosons using the iterative diagonalization approach , and was described in our earlier paper@xcite .
the diagonalizing routine which was being used for the purpose obtained all the eigenvalues and eigenvectors of the gpe , which is quite time consuming for calculations involving large basis sets . since the condensate corresponds to the lowest - energy solution of the gpe , using diagonalizing routines which obtain all its eigenvalues and eigenvectors is wasteful .
therefore , in the new version of bose_scf we now use the lapack@xcite routine dspevx to obtain the lowest eigenvalue and the eigenvector of the hamiltonian during the scf cycles , leading to substantial improvements in speed .
this subroutine aims at solving the time - independent gpe for bosons using the steepest - descent method , and was also described in our earlier paper@xcite . in this
routine , the main computational step is multiplication of a trial vector by the matrix representation of the hamiltonian . in the earlier version of the code , because the entire hamiltonian was being stored in a two - dimensional array , we used the fortran 90 intrinsic subroutine matmul for the purpose . however , now that we only store the upper triangle of the hamiltonian in a linear array , it is fruitful to use an algorithm which utilizes this aspect . therefore , we have replaced the call to matmul by a call to a routine called matmul_ut written by us .
this has also lead to significant speed improvements . as mentioned in the previous section ,
the aim of this subroutine is to multiply a vector by a real - symmetric matrix , whose upper triangle is stored in a linear array .
this routine is called from the subroutine bose_steep , and it utilizes a straightforward algorithm for achieving its goals by calling two blas@xcite functions ddot and daxpy . we have also significantly improved the capabilities of the program as far as plotting of the orbitals and the associated densities is concerned .
now the orbitals , or corresponding densities , can be computed both on one - dimensional and two - dimensional spatial grids , along user - specified directions , or planes .
the driver subroutine for the purpose is called plot_drv , which in turn calls the specific subroutines suited for the calculations
. these subroutines are plot_1d , and plot_1d_uhf for the one - dimensional plots , and plot_2d and plot_2d_uhf for the planar plots .
the output of this module is written in a file called ` orb_plot.dat ` , which can be directly used in plotting programs such as ` gnuplot ` or ` xmgrace ` .
in our earlier paper , we had described in detail how to install , compile , and run our program on various computer systems@xcite . additionally , we had explained in a step - by - step manner how to prepare the input file meant for running the code , and also the contents of a typical output file@xcite . because , various aspects associated with the installation and running of the program remain unchanged , except for some minor details , we prefer not to repeat the same discussion .
instead , we refer the reader to the readme file in connection with various details related to the installation and execution of the program . additionally , the file `input_prep.pdf` explains how to prepare a sample input file .
several sample input and output files corresponding to various example runs are also provided with the package .
in this section we report results of some of the calculations performed by our code on fermionic systems .
we present both rhf and uhf calculations for various types of traps .
further , we discuss some relevant issues related to the convergence of the calculations . in this section our aim is to investigate the convergence properties of the total hf energy of our program with respect to : ( a ) number of particles in the trap , ( b ) symmetry of the confining potential , ( c ) nature and strength of interactions , and ( d ) number of basis functions employed in the calculations . as far as the number of particles is concerned ,
we have considered two closed - shell systems namely with two particles ( @xmath72 ) , and with eight particles ( @xmath73 ) . for @xmath72 case
, calculations have been performed for all possible trap geometries ranging from a spherical trap to a completely anisotropic trap . during these calculations ,
we have considered both attractive and repulsive interactions , corresponding to negative and positive scattering lengths , respectively .
the magnitude of the scattering length ( @xmath74 ) employed in these calculations ranges from @xmath75 to @xmath76 . to put these numbers in perspective
, we recall that in most of the atomic traps , @xmath77 @xmath78 m , and for a two - component @xmath79li trapped gas , the estimated value of the scattering length is anomalously large @xmath80@xcite , where @xmath81 is the bohr radius .
thus , for this very strongly interacting system , the scattering length @xmath82 , is well within the range of the scattering lengths considered in these calculations .
therefore , the systems considered here ranging from weakly interacting ones to very strongly interacting ones truly test our numerical methods .
.convergence of total hf energy ( @xmath83 ) for a spherically symmetric trap containing two particles , with respect to the size of the basis set , for various positive values ( repulsive interactions ) of the scattering length . above
, @xmath84 is the maximum value of the quantum number of the sho basis function in a given direction , and @xmath38 is the total number of basis functions corresponding to a given value of @xmath84 . in some cases ,
fock matrix mixing approach was used to achieve convergence . [
cols="^,^,^,^,^,^",options="header " , ] finally , in fig .
[ fig : den - plot ] we present the orbital density plots for the @xmath72 case with both attractive and repulsive interactions , corresponding to @xmath85 .
the noteworthy point in the graph is the accumulation of the density at the center of the trap in case of attractive interactions , as compared to when the interactions are repulsive . with increasingly attractive interactions ,
this phenomenon becomes even more pronounced , possibly causing the instabilities in the hf approach .
plotted along the @xmath15-axis , obtained from rhf calculations on a two - particle system in an isotropic trap with @xmath86 ( solid lines ) , and @xmath87 ( dashed lines ) .
distance @xmath88 is in harmonic oscillator units .
, width=321 ] in this section we describe the results of our uhf calculations .
if one performs a uhf calculation on a closed - shell system , one must get the same results as obtained by an rhf calculation .
similarly , the total energy and orbitals of a system with @xmath5 up - spin and @xmath21 down - spin particles should be the same as that of a system with @xmath21 up - spin and @xmath5 down - spin particles .
these properties of the uhf calculations can be used to check the correctness of the underlying algorithm .
we verified these properties explicitly by : ( a ) performing uhf calculations on closed - shell systems with various scattering lengths and geometries , and found that the results always agreed with the corresponding rhf calculations , and ( b ) by performing uhf calculations on various open - shell systems with interchanged spin configurations and found the results to be identical .
therefore , we are confident of the essential correctness of our uhf program , and in what follows , we describe its applications in calculating the addition energy of fermionic atoms confined in a spherical trap .
the aim behind this calculation is to explore whether such a system follows : ( a ) shell - structure , and ( b ) hund s rule , in analogy with harmonically trapped electrons confined in a quantum dot .
we also note that a study of hund s rule for fermionic atoms confined in an optical lattice was carried out recently by krkkinen _ _ et al.__@xcite . the addition energy , _ i.e _ , the energy required to add an extra atom , to an @xmath4-atom trap is defined as @xmath89 , where @xmath90/@xmath91 represents the chemical potential of an @xmath92 particle system .
the chemical potentials , in turn , are defined as @xmath93 , where @xmath94/@xmath95 represents the total energy of an @xmath92 particle system . in our calculations ,
the total energies were calculated using the uhf approach for various values of the scattering length and our results for the addition energy for an @xmath96 spherical trap are presented in fig .
[ fig : uhf - delmu ] , for the values from @xmath97 to @xmath98 . ) of a spherical trap ( in the units of @xmath44 ) with scattering length @xmath96 , plotted as a function of the particle number @xmath4 , ranging from @xmath97 to @xmath98 .
[ fig : uhf - delmu],width=453 ] for the range of @xmath4 values studied here , in a noninteracting model the charging energy acquires nonzero values @xmath99 , only for @xmath100 @xmath101 , and @xmath102 , corresponding to filled - shell configurations . in an interacting model ,
however , @xmath103 should additionally exhibit smaller peaks at @xmath104 , @xmath105 , corresponding to the half - filled shells .
if the inter - particle repulsion is strong enough to split @xmath106 and @xmath107 shells significantly , we will additionally obtain a peak at @xmath108 corresponding to the filled @xmath107 shell , while the peaks corresponding to the half - filled shells will occur at @xmath109 , and @xmath110 , instead of @xmath105 .
moreover , it is of considerable interest to examine whether the hund s rule is also satisfied for open - shell configurations of such spherically trapped fermionic atoms , as is the case , _
e.g. _ , for electrons in quantum dots@xcite . from fig .
[ fig : uhf - delmu ] it is obvious that major peaks are located at @xmath100 @xmath101 , and @xmath102 , while the minor ones are at @xmath104 , and @xmath111 , with no peaks at @xmath109 , @xmath112 , or @xmath113 .
the heights of the major peaks are in the descending order with increasing @xmath4 , ranging from @xmath114 ( @xmath115 to @xmath116 ( @xmath117 ) .
additionally , for all the open - shell cases , the lowest - energy configurations were consistent with the hund s rule in that , a given shell is first filled with fermions of one ( say up ) spin - orientation , and upon completion , followed by the fermions of other ( down ) spin orientation .
we note that these results are qualitatively similar to the results obtained for spherical quantum dots@xcite .
thus , we conclude that for the small number of particles considered by us , the shell structure and the hund s rule are also followed by atoms confined in harmonic traps where the mutual repulsion is through short - range the contact interaction .
we have performed a number of uhf calculations on traps of different geometries , and scattering lengths , whose results will be published elsewhere .
however , we would like to briefly state that as the scattering length is increased , in several cases the ferromagnetic configurations violating the hund s rule become energetically more stable .
this implies that for large scattering lengths the uhf mean - field approach may not be representative of the true state , and inclusion of correlation effects may be necessary .
in this paper we reported a fortran 90 implementation of a harmonic oscillator basis set based approach towards obtaining the numerical solutions of both the restricted , as well as the unrestricted hartree - fock equations for spin-@xmath0 fermions confined by a harmonic potential , and interacting via pair - wise delta - function potential .
the spin-@xmath0 fermions under consideration could represent a two - component fermi gas composed of atoms confined in harmonic traps .
we performed a number of calculations assuming both attractive , and repulsive , inter - particle interactions .
as expected , the hartree - fock method becomes unstable with the increasing scattering length for attractive interactions , while no such problem is encountered for the repulsive interactions .
additionally , we performed a uhf study of atoms confined in a spherical harmonic trap and verified the existence of a shell structure , and that the hund s rule is followed .
these results are in good qualitative agreement with similar studies performed on harmonically confined electrons in quantum dots , interacting via coulomb interaction . in future , we intend to extend and improve the fermionic aspects of the present computer program in several possible ways . as far as problems related to fermionic gases in a trap are concerned
, we would like to implement the hartree - fock - bogoliubov approach to allow us to study such systems in the thermodynamic limit , and at finite temperatures . with the aim of studying the electronic structure of quantum dots , we plan to introduce the option of using the coulomb - repulsion for interparticle interactions , a step which will require significant code writing for the two - electron matrix elements .
additionally , we also aim to introduce the option of studying the dynamics of electrons in the presence of an external magnetic field , which will also allow us to study fermionic gases in rotating traps .
finally , we plan to implement the option of including spin - orbit coupling in our approach , which , at present , is a very active area of research .
we will report results along these lines in the future , as and when they become available .
e.g. _ , k. m. ohara , s. l. hemmer , m. e. gehm , s. r. granade , and , j. e. thomas , science 298 ( 2002 ) 2179 ; c. a. regal , c. ticknor , j. l. bohn , and d. s. jin , nature 424 ( 2003 ) 47 ; m. greiner , c. a. regal , and d. s. jin , 426 ( 2003 ) 537 ; s. jochim , m. bartenstein , a. altmeyer , g. hendl , s. riedl , c. chin , j. hecker denschlag , and r. grimm , science 302 ( 2003 ) 2101 ; m. w. zwierlin , c. a. stan , c. h. schunk , s. m. f. raupach , s. gupta , z. hadzibabic , and w. ketterle , phys . rev . lett . 91 ( 2003 ) 250401 ; m. w. zwierlin , a. schirotzek , c. h. schunk , and w. ketterle , science 311 ( 2006 ) 492 ; g. b. patridge , w. li , r. i. kamar , y. liao , r. g. hulet , science 311 ( 2006 ) 503 ; m. w. zwierlin , c. h. schunk , a. schirotzek , and w. ketterle , nature 442 ( 2006 ) 54 .
e. anderson , z. bai , c. bischof , s. blackford , j. demmel , j. dongarra , j. du croz , a. greenbaum , s. hammarling , a. mckenney , and d. sorensen , lapack users guide , 3rd edn . , ( 2002 ) , siam , philadelphia ( usa ) . | a set of weakly interacting spin-@xmath0 fermions , confined by a harmonic oscillator potential , and interacting with each other via a contact potential , is a model system which closely represents the physics of a dilute gas of two - component fermionic atoms confined in a magneto - optic trap . in the present work ,
our aim is to present a fortran 90 computer program which , using a basis set expansion technique , solves the hartree - fock ( hf ) equations for spin-@xmath0 fermions confined by a three - dimensional harmonic oscillator potential , and interacting with each other via pair - wise delta - function potentials .
additionally , the program can also account for those anharmonic potentials which can be expressed as a polynomial in the position operators @xmath1 @xmath2 , and @xmath3 .
both the restricted - hf ( rhf ) , and the unrestricted - hf ( uhf ) equations can be solved for a given number of fermions , with either repulsive or attractive interactions among them .
the option of uhf solutions for such systems also allows us to study possible magnetic properties of the physics of two - component confined atomic fermi gases , with imbalanced populations . using our code
we also demonstrate that such a system exhibits shell structure , and follows hund s rule .
trapped fermi gases , hartree - fock equation numerical solutions 02.70.-c , 02.70.hm , 03.75.ss , 73.21.la * program summary * + _ title of program : _ trap.x + _ catalogue identifier : _ + _ program summary url : _ + _ program obtainable from : _ cpc program library , queen s university of belfast , n. ireland + _ distribution format : _ tar.gz + _ computers : _
pcs / linux , sun ultra 10/solaris , hp alpha / tru64 , ibm / aix + _ programming language used : _ mostly fortran 90 + _ number of bytes in distributed program , including test data , etc . : _ size of the gzipped tar file 371074 bytes + _ card punching code : _ ascii + _ nature of physical problem : _ the simplest description of a spin @xmath0 trapped system at the mean field level is given by the hartree - fock method .
this program presents an efficient approach of solving these equations .
additionally , this program can solve for time - independent gross - pitaevskii and hartree - fock equations for bosonic atoms confined in a harmonic trap .
thus the combined program can handle mean - field equations for both the fermi and the bose particles .
+ _ method of solution : _ the solutions of the hartree - fock equation corresponding to the fermi systems in atomic traps are expanded as linear combinations of simple - harmonic oscillator eigenfunctions .
thus , the hartree - fock equations which comprises of a set of nonlinear integro - differential equation , is transformed into a matrix eigenvalue problem .
thereby , its solutions are obtained in a self - consistent manner , using methods of computational linear algebra .
+ _ unusual features of the program : _ none |
in the past years , there has been a growing interest in noise in mesoscopic systems @xcite .
normally , noise is an unwanted feature , and , according to classical physics , in principle can be made arbitrarily small by lowering the temperature ; according to quantum physics , however , noise is uneliminable due to the intrinsic randomness of elementary processes .
furthermore , noise , rather than being a hindrance , contains valuable information which adds to the one carried by the mean value of the quantity observed .
simple probability distributions , like e.g. the gaussian ones , are determined by the mean values and noise .
even though gaussian distributions are ubiquitous , there are interesting physical processes which are described by non - gaussian distributions .
noise alone is not sufficient for the determination of such distributions .
one needs to know all the momenta , or equivalently their generating function .
full counting statistics @xcite consists in determining the latter .
the fcs approach has been receiving increasing attention from the physics community .
its connection with the formalism of non - equilibrium green functions @xcite and circuit theory @xcite was established @xcite .
it has been used to characterize transport in heterostructures @xcite , shuttling mechanism @xcite , charge pumping @xcite , and multiple andreev reflections @xcite .
the technique was extended to charge counts in multiterminal structures @xcite , and to spin counts @xcite .
the fcs of a general quantum variable was studied , and the necessity of including the dynamics of detectors stressed @xcite .
there are some open issues in fcs .
the main one concerns whether it is always possible to find a generating function which allows an interpretation in terms of probabilities .
indeed , in ref .
it was found that such an interpretation is not straightforward .
this problem was shown to amount to the long standing question of the non - positivity of the wigner distribution @xcite .
the lack of a classical interpretation was attributed to the breaking of gauge invariance for the charge degrees of freedom , due to the presence of superconducting terminals.@xcite it is interesting to consider a more general mechanism of gauge invariance breaking which involves spin degrees of freedom .
it may be caused either by the presence of ferromagnetic terminals or by subsequent detectors measuring different components of the spin .
we shall consider the latter case .
another issue we want to address is the range of applicability of the projection postulate . since von neumann s classic work@xcite
, it is known that schrdinger s evolution can not account for the fact that the result of an individual measurement has a unique value , and can not be described by a superposition .
it is necessary to supplement schrdinger s evolution with an additional evolution ( type ii in the terminology of von neumann ) , projecting the state of the observed system into the eigenstate of the measured observable corresponding to the actual outcome .
this can be done at several stages : one could dispense with the description of the measurement , and project the wave - function of the system .
alternatively , one may continue the chain by describing the interaction of system and detector , trace out the system s degrees of freedom , and then project the state of the detector .
this chain can be continued indefinetely , by skipping the projection of the detector s state , and considering the coupling of the detector with the visible radiation , of the latter with the eye of the observer , etc .
so far , it has been implicitly assumed that the predictions of quantum mechanics do not depend on the stage at which one chooses to stop the chain and project . in this work
, we shall demonstrate that different statistics are predicted when one projects at the level of the system and at the level of the detector .
the reason for this is that , in the example we shall discuss , the quantum dynamics of the detectors can not be neglected , even after accounting for decoherence . besides being the simplest illustration of noncommuting variables , detection of spin components is a worthy subject in its own right .
spintronics , i.e. the study of how producing , detecting , and manipulating spins , is a rapidly growing field @xcite , which has already found important technological applications @xcite . in this paper , the subsequent detection of non - commuting variables is discussed .
the full counting statistics approach allows to obtain the joint probability distribution for the counts .
the non - commutativity of the observed variables manifests itself in the fact that the back - action of detectors , and their quantum dynamics , must be taken into account .
this remains true also when an environment - induced dissipative dynamics for detectors is included .
the reason is that one does not observe one particle at a time , but a flux of particles traversing the detectors at a rate which can be larger than the decoherence rate of the detectors themselves .
the present paper is laid out as follows : in section [ sec : background ] the connection of fcs with the density matrix of the detectors is derived , and a general theory of detection of non - commuting variables is presented ; a model for the measurement is introduced . in section [ sec : nodynamics ] , we discuss the case of ideal quantum detectors , having no internal dynamics .
we argue that they do not provide a realistic model of detectors because of their long memory . in section
[ sec : dynamics ] , we discuss the internal dynamics of the detectors . the fact that detectors are `` classical '' objects is accounted for by introducing a dissipative dynamics due to their interaction with an environment .
then , since we intend to concentrate on spin counts , in section [ sec : spindetector ] we present a model for a spin detector in solid state , relying on spin - orbit interaction .
we proceed to section [ sec : setup ] by introducing the particular system that we study , namely two normal reservoirs connected by a coherent conductor . in section [ sec : technique ]
we give details about the derivation of the fcs for this system , relying on the full quantum mechanical description of detection process , and we present the results . in section [ sec : pp ] , we discuss the fcs that would be obtained by a naive application of the projection postulate , i.e. by neglecting the quantum dynamics of the detectors . in section [ sec : comp12 ] , we compare the results of the two approaches for the case of one and two detectors in series , and we find that they coincide . in section [ sec : comp3 ] , we find a discrepancy between the two approaches when three detectors in series are considered . in particular , we show that both approaches predict the same second - order cross correlators , and that they differ in the prediction of fourth order cumulants @xmath0 .
finally , in section [ sec : case ] , the case of three spin detectors , monitoring the @xmath1 , @xmath2 and @xmath3 components of spin current , is presented .
the probability distribution for the counts reveals large deviations from the gaussian distribution .
all the information that we can gain about a system is stored in the density matrix of one or more detectors ( denoted by index @xmath4 ) which have interacted with the system during a time @xmath5 .
the reduced density matrix is @xmath6 where @xmath7 stands for the trace over the degrees of freedom of the measured system , @xmath8 is the initial density matrix of system and detectors , and @xmath9 is the time evolution operator .
we focus on the representation of @xmath10 in a basis @xmath11 , @xmath12 . here , @xmath13 is a vector in the hilbert space of the detectors .
since the time - evolution is linear , a matrix @xmath14 exists such that @xmath15 thus , given that one knows the initial density matrix of the detectors , @xmath16 contains all the information one can extract from the measurement .
however , part of this information gets lost : we can only know the diagonal elements of the density matrix in a particular basis , identified by the pointer states of the detectors .
these states , which will be denoted by @xmath17 , correspond to the detectors indicating the values @xmath18 , and are individuated by the property that , if one prepares the detector in a generic state identified by a density matrix @xmath19 , and then lets the environment act on it , the off - diagonal elements of the density matrix in the basis @xmath17 will go to zero with an exponential decay .
we point out that this does not dispense us from invoking a projection at some point .
the presence of the environment explains how the ensemble averaged density matrix reduces to diagonal form in the basis of pointer states , but it does not explain how the density matrix of the subensemble corresponding to an outcome @xmath20 purifies to the state @xmath21 .
this requires invoking the projection postulate for the detector , or , equivalently , an evolution dictated by the rules of the bayesian approach@xcite or of the quantum trajectory @xcite one .
the quantity accessible to observation is the probability to find the detectors in states @xmath21 , after a time @xmath5 .
it is given by @xmath22 depends only on the probabilities at a time immediately preceding @xmath5 , @xmath23 , and the process is markovian .
now , let us consider the operators @xmath24 corresponding to the read - out variables of the detectors .
their eigenstates are @xmath21 , where @xmath20 indicates an integer which is proportional to @xmath25 .
the proportionality constant is provided below .
let us also introduce the conjugated operators @xmath26 , @xmath27=i\delta_{ab}\hbar$ ] , and their eigenstates @xmath28 , with @xmath29 dimensionless quantitities proportional to @xmath30 .
if we insert to the left and to the right of @xmath10 in the rhs of eq . the identity ( in the detectors hilbert space ) in the form @xmath31 , we obtain @xmath32 } { \rho}^{\phi^-,\phi^+}_{det}(\tau)\;.\end{aligned}\ ] ] we used the shorthand @xmath33 .
we change variables according to @xmath34 . here ,
@xmath35 and @xmath36 are the classical and quantum part of the field , respectively .
this terminology reflects the fact that fluctuations of @xmath35 are set by the temperature , while fluctuations of @xmath36 depend on @xmath37 , as we shall prove in section [ sec : dynamics ] .
the time evolution depends on the hamiltonians of the system and the detectors , and on their interaction .
we focus on the detection of internal degrees of freedom of a system whose center of mass coordinate @xmath38 is not affected by the presence of the detectors .
we consider several detectors in series along the path @xmath39 .
we take the interaction to be of the form @xmath40 with @xmath41 where @xmath38 is the coordinate of the wave - packet , @xmath42 coupling constants depending on the actual detection setup , @xmath43 is an operator on the hilbert space of the system s degrees of freedom , and @xmath44 is a function which is unity inside the sensible area of the @xmath4-th detector and zero outside . for a one - dimensional motion , e.g. , we would have @xmath45 , with @xmath46 and @xmath47 are the coordinates delimiting the sensible area of the detector , @xmath48 if @xmath49 , @xmath50 if @xmath51 .
@xmath43 is the current associated with the measured quantity , such that the output of the detector does not depend on the time each particle takes to cross its sensible area .
indeed , the equation of motion for the `` measuring '' operator is @xmath52 in the equation above , we have assumed that the operator @xmath24 commutes with the unperturbed hamiltonian of the detector . in general , however , @xmath53 will fluctuate in time due to background noise .
such fluctuations put a lower limit to the resolution of the detector .
for a reliable detection , the resolution must be smaller than the minimal variation @xmath54 one intends to measure .
let us introduce proper units .
we consider the case where the measured quantities have discrete values proportional to a quantum @xmath55 .
for instance , for charge @xmath56 , the elementary charge , and for spin @xmath57 . every time an elementary unit passes the detector , the readout of the latter will change by @xmath58 .
thus , we introduce the number and phase operators @xmath59 , @xmath60 , with @xmath61 .
we further assume that _ i _ ) the detectors are initially prepared in a state with zero counts @xmath62 , and _ ii _ ) the spread of the system wave - packet is much smaller than the distance between two subsequent detectors , @xmath63 .
the first assumption implies that @xmath64 or , explicitly , @xmath65-s_{det}[\phi^-]+ \mathcal{f}_{sys}[\phi^+,\phi^-]\right)},\end{aligned}\ ] ] where the limits of the path - integrals fix the values of the fields at @xmath66 and @xmath67 , and we introduced the influence functional of the system on the detectors @xcite @xmath68\right)}:= \operatorname{tr}_{sys } \\ & { \left\ { \exp{\left(s_{int}[\phi^+,\hat{j}]\right ) } \hat{\rho}_{sys}(0 ) \exp{\left(-s_{int}[\phi^-,\hat{j}]\right ) } \right\}},\end{aligned}\ ] ] where @xmath69 is the action corresponding to the interaction @xmath70 given in eq . .
we shall call @xmath16 the quantum generating function . in principle
it depends on twice as many parameters than the classical generating function does . in the rest of the paper
we shall use the cumulant generating function ( cgf ) , @xmath71 .
the advantage of working with the cgf is that it often has a clearer interpretation than @xmath72 , since independent processes contribute factors to @xmath72 and simply additive terms to the cgf .
hence , if subsequent events are independent , the cgf is proportional to the observation time @xmath5 . thus , time averaged cumulants , which for long @xmath5 correspond to zero - frequency noise and higher order correlators for currents , have a finite value .
we analyze the situation where the dynamics of the detectors is neglected .
this means that @xmath73}=\prod_t \delta(\phi(t)-\phi),\ ] ] i.e. the counting fields are constant .
we consider first the case of one detector .
then @xmath74 , and @xmath75 where @xmath76 } \end{aligned}\ ] ] ( @xmath77 being the time - ordering operator ) is an operator in the system s hilbert space . by exploiting the cyclic property of the trace
, we have that , if @xmath78 is a conserved operator or , more generally , @xmath79=0 $ ] , then @xmath80 _ depends only on @xmath81_. it has been shown that in this case @xmath82 gives directly the generating function @xcite .
next , we consider the case of two detectors .
the kernel @xmath16 is now @xmath83 here we exploited assumption _ ii _ ) , and defined @xmath84}\;.\end{aligned}\ ] ] once again we exploit the cyclic property of the trace and see that the expression does not depend on the combination @xmath85 . from eq .
we see that in general , for two detectors , @xmath16 does depend on @xmath86 , even when @xmath43 are conserved .
however , when the system is initially in the unpolarized state @xmath87 , the dependence on @xmath88 disappears as well .
another case in which this happens is when the detectors monitor two commuting degrees of freedom which are conserved .
for instance , if the current @xmath78 is not conserved , in general @xmath89\neq 0 $ ] .
thus , even if one repeats the same measurement , one would obtain different results .
if however the current is conserved @xmath90 , and both detectors measure @xmath78 , the kernel depends only on the combination @xmath91 , which means that the two measurements will give the same result . in general , when there are three detectors , labelled 1 , 2 and 3 according to their order , measuring non - commuting quantities , even if the system is initially unpolarized , the integrand will depend on the classical variable of the middle detector , @xmath92 .
when such a dependence appears in the expression for the generating function , it is a signal that the internal dynamics of the detector must be taken into account .
indeed , when @xmath16 does not depend on @xmath35 , the density matrix is diagonal in the basis @xmath17 .
when @xmath16 does depend on @xmath35 , @xmath93 develops off - diagonal components .
we consider as an example the case where the detectors density matrix is prepared in a diagonal state at @xmath66 , and two particles are sent to the detectors one at time @xmath94 and the other at time @xmath95 , in such a way that their wave - packets do not overlap .
then , after the first particle has crossed the detectors , the density matrix of the detectors @xmath19 has offdiagonal elements , which depend on the original diagonal elements ( probabilities ) . since one observes only the probabilities , this can not be ascertained directly . however , when the second particle crosses the detectors , the new probabilities will be a combination of the former diagonal and off - diagonal elements . in order to know @xmath96 ,
knowledge of @xmath97 is not sufficient .
thus , the process is non - markovian . in principle
, even after the detector has been measuring for a long time a large number of particles , the off - diagonal elements created after the passage of the first particle will still influence its dynamics .
this is not realistic , since , because of the coupling of the detectors to the environment , the offdiagonal elements will go to zero within a typical time @xmath98 . in order to account for this
, one should consider the dynamics of the detectors , which we shall do in the next section .
we model the decoherence of the detectors by introducing a dissipative dynamics for the detectors degrees of freedom , i.e. we couple the detectors to an environment , whose degrees of freedom are traced out .
we model the environment as a system of independent harmonic oscillator in thermal equilibrium , having the action @xmath99 , \ ] ] and coupling to the detectors through the position operator @xmath100 with @xmath101 coupling constant between the @xmath102-th oscillator and the @xmath4-th detector .
then the generating function becomes @xmath103-s_{det}[\phi^-]+f_{sys}[\phi^+,\phi^-]+ s_{env}[x^+_j]-s_{env}[x^-_j ] + s_{det - env}[x^+_j,\phi^+]-s_{det - env}[x^-_j,\phi^- ] \biggr\ } \;,\end{aligned}\ ] ] in the expression above , we isolate the part @xmath104-s_{env}[x^-_j ] + s_{det - env}[x^+_j,\phi^+]\\ & \phantom{s_{det - env}}-s_{det - env}[x^-_j,\phi^-]\biggr ) \rho_{env}(x^+,x^-)\end{aligned}\ ] ] which gives the influence functional of the environment on the detectors .
we notice from eq . that , since the functions @xmath105 are fixed by the external path - integrals , they act as an external source @xmath106 on the @xmath102-th harmonic oscillator .
it is then possible to perform the independent gaussian path - integrals over @xmath107 , resulting in@xcite @xmath108\;,\end{aligned}\ ] ] where the influence of the environment is contained in the complex functions @xmath109 , whose fourier transforms are @xmath110 where the inverse temperature @xmath111 comes from having assumed the bath in thermal equilibrium ( @xmath112 ) , and @xmath113 are the spectral densities @xmath114\;,\end{aligned}\ ] ] at low frequencies , we can approximate the odd - functions @xmath113 by @xmath115 ( ohmic approximation ) , with @xmath116 friction constant , as will be clear later .
we introduce new variables @xmath81 , @xmath117 .
thus we get @xmath118 we take the action of free detectors to be that of harmonic oscillators , i.e. @xmath119=\sum_a \frac{-i m_a v_{qa}^2}{2\hbar}\int \frac{d\omega}{2\pi } ( \omega^2-\omega_a^2 ) \left|\phi_a(\omega)\right|^2\;,\ ] ] where @xmath120 is the mass " of the detector ( i.e. it is the inertial term corresponding to the kinetic energy @xmath121 )
. then the generating function reads @xmath122 + \mathcal{f}_{sys}[\phi,\phi]\biggr\ } \;,\end{aligned}\ ] ] where we introduced the response function , @xmath123 from which one can see that @xmath116 are proportional to the friction constant , and the fluctuation term @xmath124 the part of the action containing the fluctuation term in @xmath125 is , at low frequencies , proportional to temperature @xmath126 and to @xmath127 .
the factor @xmath127 strongly suppresses large fluctuations in @xmath36 .
thus , the influence functional due to the measured system @xmath128 = \int dt l_{inf}(\phi(t),\phi(t))$ ] can be approximated by @xmath129 $ ] .
integration over @xmath130 gives finally @xmath131+s_{eff}[\phi]}\;,\end{aligned}\ ] ] with the effective action @xmath132= -\frac{1}{2 } \sum_a \frac{(\beta m_a v^2_{qa})^2}{\gamma_a } \int d\omega \frac{|g^{-1}_a(\omega)|^2}{f(\omega ) } \left|\phi_{a}(\omega)\right|^2 \;.\ ] ] we notice that at high temperatures @xmath133 , and thus @xmath37 disappears in the effective action for @xmath35 .
for this reason the latter is termed the `` classical '' part of the field . in the limit of small mass @xmath134 , @xmath135 , where @xmath136 has a finite value and is a typical energy scale of detector @xmath4
, the effective action simplifies to @xmath132= -\frac{1}{2}\sum_a\int dt \left [ \tau_{ac } \left(\dot{\phi}_{a}(t)\right)^2 + \frac{1}{\tau_{ac } \delta \phi_a^2}\phi_a(t)^2\right ] \;,\ ] ] with @xmath137 the `` coherence time '' of the detector , and @xmath138 the spread of @xmath139 .
we discuss a model for spin detection .
the setup corresponds to the one proposed and used in @xcite to detect aharonov - casher effect @xcite for neutrons .
this setup exploits the fact that a moving magnetic dipole generates an electric one @xcite . to measure this ,
one encloses the two - dimensional current lead between the plates of a capacitor as shown in fig .
[ fig : detector ] . while in ref .
the neutrons passed a fixed electric field , which gave a constant aharonov - casher phase , in a spin detector the initial voltage applied to the plates is zero , and the passing of a particle with spin 1/2 will cause the charge in the capacitor to show pulses towards positive or negative values depending on the result of the measurement .
the associated phase @xmath140 will thus increase or decrease stepwise in the ideal situation where spins are transmitted separately in vacuum through the detector .
each spin moving with velocity @xmath141 produces an electric field . for electrons in vacuum
, the interaction term between spin and detector is given by the spin - orbit coupling @xmath142 with @xmath143 the speed of light , and the factor @xmath144 accounts for the thomas precession .
the magnetic moment @xmath145 is proportional to the spin @xmath146 , with @xmath147 mass of the electron , @xmath148 its charge , and @xmath149 its spin gyromagnetic factor .
thus , we rewrite the interaction as @xmath150 the spin - orbit coupling induces a current in the @xmath151 circuit . the integrated charge traversing
the circuit is the detector read - out .
the read - out signal is proportional to spin current in the lead @xmath152 , @xmath153 , @xmath154 being the unit vector perpendicular to the direction of the current flow and parallel to the plates of the capacitor , @xmath155 being a proportionality coefficient .
the concrete expression for the latter , @xmath156 , depends on the geometrical dimensions of the detector the length of its plates in the direction of the current @xmath157 , and the distance between the plates @xmath158 . the proposed spin current detector .
an electron with velocity @xmath141 and spin @xmath159 induces a voltage drop in a capacitor .
the electric field @xmath160 inside the capacitor produces an aharonov - casher phase shift on the electrons.,scaledwidth=40.0% ] the variable canonically conjugated to the read - out is the voltage @xmath161 across the capacitor , and the expression for the interaction in terms of @xmath161 contains the same proportionality coefficient @xmath155 , @xmath162 .
our choice of the detection setup is motivated by the fact that this detector does not influence electron transfers through the contact and only gives the minimal feedback compatible with the uncertainty principle : the electrons passing the capacitor in the direction of current acquire an aharonov - casher phase shift , which consists in a precession of the spin around the detecton axis @xmath154 .
this depends on spin and is given by @xmath163 .
this is similar to the detection scheme presented in @xcite for charges transferred .
a fundamental complication in comparison with the charge fcs is that in our case the phase shift depends on spin , so that even the minimal feedback influences the statistics of the outcomes of following spin detectors .
we introduce dimensionless variables @xmath164 , @xmath165
. then @xmath166 varies by one every time a spin 1/2 crosses the detector . with reference to eqs .
, , we have @xmath167 , @xmath168 , @xmath169 , @xmath170 , @xmath171 , with @xmath172 , @xmath173 and @xmath174 inductance ( assumed negligible ) , resistance and capacitance of the circuit .
we consider a system composed of two metallic , unpolarized leads , connected through a coherent conductor , characterized by a set of transmission probabilities @xmath175 , where @xmath176 identifies transmission channels .
a negative bias voltage @xmath161 is applied to the left lead . at the right of the conductor
there are several spin detectors , labelled from left to right by @xmath177 , and a current detector , denoted by @xmath178 .
the counting fields will be then @xmath29 , with @xmath179 .
since charge and spin currents commute , the current detector can be positioned at any point along the chain of detectors , without influencing the statistics of the outcomes .
the setup is depicted in fig .
[ fig : setup ] .
the setup considered , in the case of three spin detectors and one charge detector.,scaledwidth=30.0% ] we require that the coherent conductor is non - polarizing .
thus , the average spin current is zero .
however , there are spin fluctuations , which are revealed by measuring noise and higher order correlators ( or cumulants ) . from the symmetry with respect to reversal of spin
, we can predict a priori that all odd cumulants are zero .
we shall concentrate on a situation where there are three spin detectors .
this is because , as anticipated in section [ sec : background ] , the current is unpolarized and one needs at least three detectors monitoring non - commuting quantities in order to see non - trivial consequences of the detectors feedback on the system .
the feedback consists in the wave - function picking up an aharonov - casher phase while traversing each detector .
the technique we use is an extension of the scattering theory of charge fcs .
this theory @xcite expresses fcs in terms of a phase factor @xmath180 acquired by scattering waves upon traversing the charge detector .
since we do not consider energy - resolved measurements , the phase factor does not depend on the channel , and the approach works for a multi - channel conductor as well as for a single - channel one .
the phase factor @xmath180 can be seen as resulting from a gauge transform , to be applied to the ( known ) green function of the right lead , that removes the coupling term @xcite @xmath181 . for the case of the spin detectors ,
the gauge transform introduces a phase factor which is a unitary matrix in spin space .
namely , the gauge transform generated by spin detector @xmath4 is @xmath182 . in this matrix
, @xmath183 is a pseudovector of @xmath184 pauli matrices , and @xmath185 is the direction along which detector @xmath4 detects spin current .
the keldysh green function of the lead is @xmath186 where @xmath187 denotes the left or right lead , and @xmath188 is the corresponding fermi occupation number at energy @xmath189 and chemical potential @xmath190 .
the elements of the matrix are actually in their turn a matrix in spin space . since the leads are assumed to be unpolarized , they are simply the identity
. the matrix current is given by @xcite @xmath191 } { 1 + { t}_{n } \left(\left\ { \check{g}_{l},\check{\tilde{g}}_{r}\right\ } -2\right)/4}\;,\ ] ] from which it follows that the quantum generating function is @xmath192 here @xmath193(\{ ...
\})$ ] denote ( anti)commutator of two matrices , and @xmath194 is the transformed matrix @xmath195 where @xmath196 , @xmath197 are matrices in keldysh space . after substituting the expression for @xmath198
, we obtain @xmath199 where @xmath200 is a matrix in spin space .
we notice that _ i _ ) the charge fields come only in the combination @xmath201 , and _ ii _ ) the phase factors @xmath202 in the expression for charge fcs are replaced by @xmath203 to give the fcs of charge and spin counts after taking trace over spin .
if we also notice that @xmath204 is a @xmath205 matrix with eigenvalues @xmath206 , we arrive to @xmath207 where @xmath208 is the generating function for charge counting .
the @xmath209 is given by @xmath210 .
the explicit expression for the system considered here , in terms of the transmission probabilities through the contact and the applied bias is , at zero temperature , @xmath211}\ ; , \label{eq : main_relation2}\ ] ] with @xmath212 , @xmath213 .
the interpretation is quite straightforward : electrons coming through different channels behave independently , which is revealed by the fact that the generating function splits into a sum ; each channel can accommodate two electrons in a spin - singlet configuration ; with probability @xmath214 none of the two electrons passes the junction , and there is no contribution to the charge counting nor to the spin one ; with probability @xmath215 both electrons come through the conductor .
this gives a contribution of two elementary charges transferred ( factor @xmath216 ) , but no spin transfer .
finally , with probability @xmath217 , exactly one of the two electrons is transferred .
this gives a contribution to the charge and to the spin counting .
we demonstrate that a different fcs is predicted by using a different approach , namely a nave application of the projection postulate , consisting in avoiding the description of the measurement and applying the projection to the system measured .
we shall denote this procedure with pp for brevity .
this approach predicts a parameter @xmath218 which does not depend on @xmath35 .
let us give the details of such a derivation : when an unpolarized electron arrives to the first detector , the probability of the outcome @xmath219 is @xmath220 .
the conditional probability that the second detector gives @xmath221 , given that the first read @xmath222 is @xmath223 .
this is because after the first detection the spin of the electron is assumed to have collapsed along @xmath224 .
the same happens after the second detection .
consequently , the conditional probability that a third detector reads @xmath225 , given that the first read @xmath222 and the second @xmath221 , depends only on the latter outcome @xmath226 .
the process is in a sense a markovian one .
the total joint probability for each electron transfer with an arbitrarily long chain of detectors is @xmath227 and the corresponding generating function for the setup considered here is given by eq . with @xmath228 }
\frac{1}{2 } \prod_{a=1}^{k-1 } \frac{1\!+\!\sigma_{a } \sigma_{a+1 } \mathbf{n}_a\!\cdot\!\mathbf{n}_{a+1}}{2 } \;.\end{aligned}\ ] ]
now , let us go back to eqs . , and compare the two approaches for some simple cases . for the case of one or two detectors in series , the eigenvalues @xmath229 are not affected by the order of matrix multiplication in ( [ eq : matrix ] ) and depend on differences of spin counting fields @xmath230 only ( in fact they coincide with the value @xmath231 ) .
this implies that the fcs definition ( [ eq : fcs_definition ] ) can be readily interpreted in classical terms : it is a generating function for probability distribution of a certain number of spin counts @xmath113 in each detector , @xmath232 for a single detector , the spin fcs is very simple : it corresponds to independent transfers of two sorts of electrons , with spins `` up '' and `` down '' with respect to the quantization axis .
the cumulants of the spin ( charge ) transferred are given by the derivatives of @xmath233 with respect to @xmath234 ( @xmath235 ) , at @xmath236 . in this case
@xmath237 . from this and relation ( [ eq : main_relation ] ) , we conclude that all odd cumulants of spin current are 0 , as anticipated , and all even cumulants coincide with the charge cumulants . for two spin detectors , with @xmath238 , we obtain @xmath239 .
since there is no dependence on @xmath139 , the quantum generating function has an immediate interpretation ; we consider the case when the read - out of the charge is not exploited ( @xmath240 )
. then @xmath241^m\!,\ ] ] where @xmath242 is the probability that , in two attempts of transmitting one electron over a spin degenerate channel @xmath176 , exactly one is transmitted and @xmath243 .
this result coincides with what one would obtain from the projection postulate .
we discuss in detail the probability distribution . by performing the fourier transform
, we find the probability of detecting a spin @xmath222 in direction @xmath244 and @xmath221 in direction @xmath245 : @xmath246 where the prime in the sum means that it is restricted to @xmath247 , and the probability for each channel @xmath176 is @xmath248 where @xmath249^{l } [ \sin^2{(\theta/2)}]^{k - l}\;,\\ p_\up(l|k=\dwn)=&\ \binom{k}{l } [ \cos^2{(\theta/2)}]^{k - l } [ \sin^2{(\theta/2)}]^{l}\end{aligned}\ ] ] the sums are over all values for which the binomials make sense ( no negative nor half - integer values ) .
thus @xmath250 have the same parity .
@xmath251 can be interpreted as follows : since the current is unpolarized , we can think of it as carried by pairs of electrons in singlet configuration .
then , there is a successful attempt to transfer spin when exactly one of the two electrons is transmitted .
this gives @xmath252 , the probability of transferring @xmath253 spins over @xmath166 attempts ( @xmath254 probability of success for a single attempt ) through channel @xmath176 ; the second binomial comes from the ways one can pick @xmath255 spins up out of @xmath253 spins , with probability @xmath144 ( we recall that the incoming electrons are unpolarized ) ; the third term comes from the fact that , given that @xmath255 spins up according to the first detector are passed to the second one , the latter will measure @xmath256 of these as spins up ( the probability of agreement between detectors being @xmath257 ) , and the rest as spins down ; analogously , the latter term comes from the fact that given @xmath258 spins down along direction @xmath244 have been detected , @xmath259 of them will be detected from the second detector as spins up , while the remaining ones will be detected as down . when the two detectors have parallel orientation ( @xmath260 ) , the second sum in eq .
is nonzero only if @xmath261 , giving @xmath262 i.e. there is perfect correlation , as is to be expected .
when the two detectors have orthogonal orientation ( @xmath263 ) , it is possible to perform analytically the sum over @xmath264 : @xmath265 i.e. the outcomes are independent , given that @xmath253 successful spin transfers happened .
for the case of three detectors , we have @xmath266 here @xmath267 are the angles between the polarizations ( @xmath154 ) of detectors @xmath102 and @xmath253 , and @xmath268 , @xmath269 . as
before @xmath270 is the part corresponding to the projection postulate .
we notice that when two consecutive detectors are parallel or antiparallel , then @xmath271 .
this is because the same measurement is repeated twice , and thus we fall back to the case of two detectors . in general , however , @xmath272 depends on @xmath92 , and thus one needs to account for the dynamics of the second detector in order to get the probability distribution for the spin counts .
we recall that the corresponding detector s action is @xmath273=\int dt \frac{1}{2}\left[\tau_c \dot{\phi}_2(t)^2 - \phi_2(t)^2/\tau_c \langle\phi_2 ^ 2 \rangle\right]$ ] , with @xmath98 coherence time and @xmath274 fluctuations of @xmath92 .
we have calculated the second cumulants or cross - correlators : we found that they differ from the ones obtained by using pp only by small terms .
the correlator between first and third detector s readings is : @xmath275 , \end{aligned}\ ] ] where @xmath276 , and the first term is the pp result .
the second term , as expected , has a typical signature of interference effects : it is suppressed exponentially if the variance of the corresponding aharonov - casher phase @xmath277 . since @xmath278 is inversely proportional to @xmath37 ,
this is the classical limit . in this limit ,
the result coincides with the pp . however , fourth cumulants show a large deviations from the pp result .
namely : @xmath279 where and the pp result is expressed in terms of charge cumulants as @xmath280 .\ ] ] this deviation results from correlations of @xmath92 at time scale @xmath98 . to estimate the result , we notice that the charge cumulants are of the order of @xmath281 , @xmath282 being the average time between electron transfers .
it is easy to fulfill the condition @xmath283 , and in this case @xmath284 is much larger than pp result .
it is interesting to study further the probability distribution which gives rise to such anomalously large fourth - order cumulants .
this we shall do in the next section .
we discuss for definiteness the case of three detectors oriented along three orthogonal directions forming a right - handed basis .
this implies @xmath285 .
we concentrate on the joint probability distribution for the outcomes of the first and the third detector , irrespectively of the reading of the second detector .
we consider the `` classical '' limit @xmath286 .
then , the generating function for the probability @xmath287 for counting @xmath288 spins in the detectors is : @xmath289}\right ] \biggr\}\end{aligned}\ ] ] where @xmath290 we have a path - integral over imaginary time .
we exploit the quantum mechanical technique and re - express the path - integral in terms of amplitudes : @xmath291 here the counting fields @xmath36 are parameters , and the time evolution of the variable @xmath92 is dictated by @xmath292 , with the hamiltonian @xmath293 } \;.\ ] ] then , for large values of @xmath5 , the path - integral can be approximated @xmath294 where @xmath295 is the ground state energy of the hamiltonian .
the next step is to find an explicit expression for the probability .
we recall that the probability to have detectors 1 and 3 measure average spin currents @xmath296 , @xmath297 is related to @xmath298 through @xmath299 since we are in the large @xmath5 limit , we can evaluate the integrals in the saddle - point approximation , and obtain @xmath300}\;,\ ] ] where @xmath301 satisfy the saddle point condition [ eq : sp ] @xmath302 assuming that the solutions are much smaller than 1 , @xmath303 , we have that the hamiltonian can be rewritten , including terms up to second order in @xmath36 , as @xmath304 \;,\ ] ] where we introduced the average time between spin transfers , @xmath305 .
we recognize the hamiltonian for the mathieu equation @xmath306 thus the ground state energy depends on the lowest mathieu characteristic function @xmath307 , with the coupling strength given by @xmath308 .
namely , @xmath309 the saddle - point equations can then be combined to give a trascendent equation for @xmath310 , from which one expresses @xmath311 , which are purely imaginary , according to [ eq : spsol ] @xmath312 here , @xmath313 is the solution to the equation @xmath314 ^ 2 } + \frac{(\nu_1-\nu_3)^2}{\left[2-a'_0(q)\right]^2 } \;,\ ] ] where we introduced dimensionless currents @xmath315 .
eqs . , are valid in the limit @xmath316 , i.e. @xmath317 . finally , we have that the probability distribution is @xmath318 ^ 2 \\ & \label{eq : prob } - ( \nu_1-\nu_3)^2 ( 1-a'_0(q^*))/ \left[2-a'_0(q^*)\right]^2\;. \end{aligned}\ ] ] this probability distribution is to be compared with the one predicted by applying the pp .
the latter is , in the same regime @xmath316 , the independent combination of two gaussians : @xmath319 the proportionality constant ( @xmath320 ) being the same . in the limit @xmath321 , we have that eqs . and
however , by taking into account that the detectors have a finite decoherence time @xmath98 , and that the time between spin transfers @xmath322 can be much smaller than @xmath98 , we find that the probability distribution deviates sensibly from eq .. this deviation is larger in the regime @xmath323 . when both ( dimensionless ) currents are comparable in module and large with respect to 1 . when @xmath324 , we find that @xmath325 with the scaling function @xmath326 defined by @xmath327 where the condition @xmath328 defines @xmath329
. in particular , @xmath326 diverges at @xmath330 according to @xmath331 . the log of probability as a function of @xmath332 for different values of @xmath333 for the configuration studied in the text .
all the curves have been shifted by @xmath334 .
the upper curves correspond to the result of the fcs approach , and the lower ones to the pp .
the black dotted curve is the limiting scaling curve discussed in the text.,scaledwidth=40.0% ] in fig .
[ fig : comp ] we draw the logarithm of probability as a function of @xmath332 for several values of @xmath333 , and compare with the probability predicted by making use of pp .
we have discussed the full counting statistics of non - commuting variables . as a concrete example , we focused on spin counts in a two terminal device with non - ferromagnetic leads connected through a non - polarizing coherent conductor .
we have provided a formula connecting the fcs of spins to the one of charge .
we have seen that it is crucial to have a coherent conductor with finite transparency connecting the two leads .
this is because electrons transmitted through the same channel are in a spin singlet , and thus contribute no net spin transfer nor spin fluctuations .
however , if the transmission probability through channel @xmath176 is finite ( @xmath335 ) , then there is a non - zero probability @xmath336 that exactly one electron out of a singlet pair is transmitted , and this contributes to spin fluctuations .
another interesting conclusion which we can draw from this work is that , when measuring non - commuting quantities with subsequent detectors , one should take into account the quantum dynamics of the detectors themselves .
this is because the decoherence time for the detectors , @xmath98 can be larger than the average time between two subsequent counts , @xmath322 . thus if one would put by hand the off - diagonal elements of the detectors density matrix to zero after each count , which amounts to applying the projection postulate , one would obtain the wrong result .
we have shown that , in the system considered here , such a deviation from the nave application of the projection postulate is revealed by the fourth correlator of spin counts .
we acknowledge the financial support provided through the european community s research training networks programme under contract hprn - ct-2002 - 00302 , spintronics . | we discuss the full counting statistics of non - commuting variables with the measurement of successive spin counts in non - collinear directions taken as an example .
we show that owing to an irreducible detector back - action , the fcs in this case may be sensitive to the dynamics of the detectors , and may differ from the predictions obtained with using a naive version of the projection postulate .
we present here a general model of detector dynamics and path - integral approach to the evaluation of fcs .
we concentrate further on a simple diffusive " model of the detector dynamics where the fcs can be evaluated with transfer - matrix method .
the resulting probability distribution of spin counts is characterized by anomalously large higher cumulants and substantially deviates from gaussian statistics . |
diffractive processes contribute a significant fraction to the total inelastic proton - proton cross sections at high energies .
these reactions can be described in terms of the exchange of a @xmath3 , a hypothetical object with the quantum numbers of the vacuum .
the experimental signatures of diffractive events are the presence of non - exponentially suppressed large rapidity gaps and/or presence of the intact leading protons .
diffractive events with a hard parton - parton scattering , so called _ hard diffractive events _ , subject of these proceedings , are of particular interest since they can be studied in terms of perturbative qcd . the measurements presented here are based on the data collected by the cms experiment during 2010 at a @xmath47 tev .
the detailed description of the cms experiment can be found elsewhere @xcite .
the simulation of non - diffractive events is obtained with pythia6 @xcite and pythia8 @xcite .
hard diffractive events are simulated with the pompyt @xcite and pomwig @xcite generators , as well as pythia8 , these generators are used with diffractive pdfs ( dpdfs ) from the same fit to diffractive deep inelastic scattering data , h1 fit b.
dijet events were selected by requiring 2 jets with @xmath520 gev and [email protected] . the anti-@xmath7 clustering algorithm with the radius parameter @xmath8=0.5
was used to reconstruct jets . to enhance the diffractive contribution ,
the requirements @xmath93 ( @xmath10 - 3 ) were enforced , where @xmath11 is the pseudorapidity of the most forward ( backward ) particle - flow ( pf ) @xcite object , combining measurement from the tracker and the calorimeters .
this selection translates to imposing a rapidity gap of at least 1.9 units in the hadron calorimeter ( hf ) ( with the coverage [email protected] ) .
figure [ diff - dijet](a ) shows the effect of this selection , the @xmath13 requirement rejects events at high @xmath2 values , whereas region of low @xmath2 dominated by the diffractive contribution is not affected .
the results are compared to mc predictions where the relative diffractive contribution was scaled by a factor of 0.23 obtained by minimizing the difference between @xmath2 distributions of the data and sum of non - diffractive and diffractive models .
the diffractive cross section as a function of @xmath2 is shown in fig .
[ diff - dijet](b ) .
the low @xmath2 bin shows significant contribution from diffractive dijet production , observed for the first time at the lhc @xcite . the associated _ rapidity gap survival _
probability is estimated to be in the range from distributions with ( open symbols ) and without ( closed symbols ) the @xmath9 3 requirement , compared to mc predictions ; ( b ) the differential cross section for inclusive dijet production as a function of @xmath2.,title="fig : " ] distributions with ( open symbols ) and without ( closed symbols ) the @xmath9 3 requirement , compared to mc predictions ; ( b ) the differential cross section for inclusive dijet production as a function of @xmath2.,title="fig : " ] [email protected] ( nlo ) to [email protected](lo ) by comparing measured cross section and predictions from the leading order ( pompyt and pomwig ) and next - to - leading order ( powheg ) @xcite diffractive generators based on dpdfs from the hera experiments .
the identification of w and z events is based on the presence of isolated electrons and muons with high transverse momentum .
electron ( muon ) candidates are required to have @xmath151.4 and @xmath525 gev , while events with a second isolated electron or muon with @xmath510 gev are rejected , missing transverse momentum is required to be greater than 30 gev and the transverse mass of the charged lepton and the neutrino to be greater than 60 gev ; similarly for z candidates leptons were required to have @xmath525 gev and @xmath161.4 with the reconstructed invariant mass of the dilepton system restricted between 60 and 120 gev .
this selection results in essentially background - free w and z event samples .
[ diff - w](a ) shows charged particle multiplicity for w events ( electron and muon channels combined ) with a large rapidity gap ( lrg ) signature , enforced by requiring that none of the calorimeter towers had a measured energy of more than 4 gev in at least one of the hf calorimeters , which corresponds to a pseudorapidity interval of 1.9 units .
the data are compared with the predictions of pythia6 and pythia8 .
events with zero energy deposition manifest the presence of a pseudorapidity gap extending over hf .
however large discrepancies between the data and different models are observed .
the percentage of w and z events with lrg signature is ( [email protected](stat.)@xmath140.38(syst.))% and ( [email protected](stat.)@xmath140.42(syst.))% .
a large asymmetry is observed between the number of events with the charged lepton from w decay in the opposite and with the same hemisphere as the rapidity gap , in accordance with the predictions of pompyt diffractive mc , see fig .
[ diff - w](b ) . in comparison ,
the various non - diffractive mc predictions predict a symmetric distribution , see pythia curves on fig .
[ diff - w](b ) .
the diffractive component is estimated to be ( [email protected](stat.)@xmath145.2(syst.))% by fitting the observed asymmetry with the admixture of pompyt and non - diffractive ( pythia ) events .
this presents first evidence for diffractive w production at the lhc @xcite .
i would like to thank the organizers of 15th international conference on elastic & diffractive scattering for warm hospitality and for an exciting conference .
99 s. chatrchyan _
et al . _
( cms collaboration ) , journal of instrumentation * 3 * , s08004 ( 2008 ) .
t. sjostrand , s. mrenna and p. skands , jhep * 0605 * ( 2006 ) 026 .
t. sjostrand , s. mrenna and p. skands , comput.phys . commun .
* 178 * ( 2008 ) 852 .
p. bruni and g. ingelman , phys .
b * 311 * , 317 ( 1993 ) .
cox and j.r .
forshaw , comput.phys.commun .
* 144 * ( 2002 ) 104 .
( cms collaboration ) , _ cms physics : technical design report volume 1 : detector performance and software _ , tech.rep .
( cern lhc , 2006 ) . s. chatrchyan _
( cms collaboration ) , phys .
d 87 ( 2013 ) 012006 s. alioli _ et al . _ ,
jhep * 04 * ( 2011 ) 081 . s. chatrchyan _
( cms collaboration ) , epjc 72 ( 2012 ) 1839 | in these proceedings we present cms results on hard diffraction .
diffractive dijet production in @xmath0 collisions at @xmath1=7 tev is discussed .
the cross section for dijet production is presented as a function of @xmath2 , representing the fractional momentum loss of the scattered proton in single - diffractive events .
the observation of w and z boson production in events with a large pseudo - rapidity gap is also presented . |
galaxies are intrinsically complex with multiple components and varied formation histories .
this complexity is the primary reason that unravelling the physics of galaxy formation and evolution is so challenging .
galaxies are made up of baryons confined to dark matter haloes , and often have multiple distinct kinematic components ( e.g.bulge and/or disc ) .
there are complex interactions between the stars , gas , dust , dark matter and super - massive black holes .
these can lead to both positive and negative feedback on the formation rate of stars .
experimental efforts to address galaxy formation have generally taken two directions .
first , galaxy imaging and spectroscopic surveys have progressively moved to higher redshift , in an attempt to directly observe galaxy evolution and formation .
this approach has had much success , placing strong constraints on the evolution of the global star formation rate ( e.g. * ? ? ?
* ) , unveiling the strong evolution of black hole accretion over most of cosmic time ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , tracing the evolution of galaxy size and morphology ( e.g. * ? ? ? * ) and much more .
the second approach has been to expand our view in wavelength rather than cosmic time .
the physical processes occurring in galaxies cause emission over the entire range of the electromagnetic spectrum . in order to have a full picture of galaxy formation
, a multi - wavelength approach is vital .
this has been made more achievable with recent generations of satellites covering the x - ray , ultra - violet , mid- and far - infrared .
while the spectral energy distributions of stars tend to peak in the optical or near - infrared , obscuration and reprocessing by dust generates strong mid- and far - infrared emission .
young stars ( when not obscured ) are most directly traced in the ultraviolet , while black hole accretion can generate radiation from the radio to x - ray and gamma - ray bands .
surveys such as the galaxy and mass assembly ( gama ) survey @xcite and the cosmic evolution survey ( cosmos ; * ? ? ?
* ) show the value of this multi - wavelength approach . the third route , and the one that we address in this paper , is to focus on spatially resolving galaxies ; in particular , obtaining spatially resolved spectroscopy .
optical spectroscopy allows us to measure a wide range of parameters including current star formation rates ( e.g. via ) , gas phase metallicities , stellar ages , stellar metallicities , black hole accretion rates , ionization structure and extinction due to dust ( e.g. via the balmer decrement ) . the major spectroscopic surveys to date have used a single fibre @xcite or single slit on each object , and so obtain just one measurement of these parameters for each galaxy . moreover , these measurements may not be representative of the galaxy as a whole , but biased , depending on where the aperture is placed .
this fundamental problem is addressed by integral field spectrographs ( ifs ) . in the last decade
, projects such as sauron @xcite have demonstrated the power of integral field spectroscopy to capture a range of key observables that are simply not available to single - aperture surveys . as well as studying the properties listed above in a spatially - resolved context , obtaining gas and stellar kinematics over an entire galaxy enables us to separate dynamical components , measure dynamical mass , examine the impact of winds and outflows , and
discover merging systems via dynamical disturbances .
integral field spectroscopy has almost exclusively been limited to single - object instruments , meaning that it is time - consuming to build large samples .
the largest current data set , using the sauron system on the william herschel telescope , contain @xmath3 objects ( atlas-3d ; * ? ? ?
* ) . the califa project @xcite aims to target 600 objects with the pmas integral field unit ( ifu ) on the calar alto telescope using @xmath4200 nights of telescope time . the only multi - object integral field spectrograph currently available is that on the vlt flames instrument @xcite , which contains 15 ifus , each with 20 spatial elements of size 0.52 arcsec and a total field of view ( fov ) of @xmath5arcsec .
this has enabled measurements of the tully - fisher relation at @xmath40.6 as well as a number of other projects .
however the small fov of each ifu , combined with the high spectral resolution ( @xmath69000 ) and associated narrow wavelength range limits its applicability for large scale surveys .
astrophotonic technology @xcite is now opening the way to new instrumentation that can address the need for highly multiplexed integral field spectroscopy .
hexabundles @xcite are optical fibre bundles where the cladding has been stripped from each fibre to a minimum over a short length ( @xmath430 mm ) and the fibres then gently fused together at the input end to provide an ifu ( @xmath41 mm aperture ) with high filling factor . these can then be used in conventional multi - fibre spectrographs . in this paper ,
we report on the sydney - aao multi - object integral field spectrograph ( sami ) , the first fully operational demonstrator instrument to use hexabundles .
sami has 13 ifus that can be positioned anywhere over a 1 degree diameter field of view . in section [ sec : rationale ] , we discuss in detail the scientific rationale for such an instrument , along with some practical considerations regarding sensitivity . in section [ sec : instrument ] , we describe the sami instrument in detail . in section [ sec : comm ] , we outline the observations carried out during the commissioning of the instrument , the results from which are discussed in section [ sec : results ] . our conclusions , and goals for the future ,
are laid out in section [ sec : conc ] .
in this section we outline the key scientific drivers for an instrument such as sami .
the fundamental question at the heart of this work is : how did the galaxy population we see around us today come about ?
this requires us to understand the physical processes that occur as galaxies form and evolve .
the galaxy population we see today has some very distinctive features that need to be explained .
one of the most fundamental is the separation of galaxies into a bimodal distribution according to colour ( e.g. * ? ? ?
* ; * ? ? ?
a galaxy s colour is primarily related to the presence of ongoing star formation .
the second key feature differentiating galaxies is morphology .
there is a strong correlation between colour and morphology , with galaxies lying along the ` red sequence ' being mostly passive systems with elliptical / spheroidal morphology , while galaxies inhabiting the ` blue cloud ' are mostly dominated by discs , although this is not always strictly the case ( e.g. * ? ? ?
* ; * ? ? ?
while they are related , there is not a strict one - to - one relationship between morphology and colour .
a clearer understanding of galaxies can be obtained if they are considered as being made up of distinct morphological components ( discs , bulges and pseudo - bulges ) that result from different formation processes and evolutionary histories @xcite .
the intrinsic properties of these structural components are more uniform than those of the galaxies they compose .
their formation pathways are also quite different , with true bulges built up by violent mergers , discs from gas accretion , and pseudo - bulges from secular evolution .
disentangling these various modes is complex , but can be materially aided by the fact that the structures formed have different kinematic properties as well as different star formation histories . in order to ascertain the physics involved ,
we need to determine the answers to a number of questions that broadly fall into four categories : ( i ) how does galaxy mass and angular momentum build up ? ( ii ) when , where and why does star formation occur ?
( iii ) when , where and why does black hole accretion occur ? ( iv ) how are galaxies fuelled and what is the role of feedback
? we will discuss each of these issues in turn , although there is of course significant overlap between them .
the standard picture of galaxy formation has gas cooling to form a rotationally supported disc within a cold dark matter halo @xcite .
while this picture is broadly accepted , feedback and interactions provide major complications which are not yet fully understood .
the scaling relation between circular velocity and stellar mass for disc galaxies provides a tight constraint for galaxy formation models .
the circular velocity depends on the ratio of disc mass to halo mass , the dark matter halo profile and the dimensionless spin parameter , @xmath7 @xcite .
the largest tfr samples ( e.g. * ? ? ?
* ) currently contain @xmath8 galaxies with either long - slit spectroscopy or spatially unresolved velocities .
if ifu observations can be made out to large enough radii ( typically @xmath9 disc scale lengths ) , then they provide substantial advantages in allowing a clearer picture of distorted kinematics and inclination .
circular velocities can be compared to the results of galaxy lensing to constrain the dark matter halo profile and examine evidence of contraction of the halo in response to the baryons @xcite .
the stellar and emission line kinematic data that sami can provide will allow dynamical mass estimates within maximum radius probed , using techniques such as anisotropic jeans modelling @xcite .
in general it is not possible to determine total mass because of the uncertainty of the dark matter halo parameters .
even for the milky way , where many halo stars can be used to probe the outer halo , the total galaxy mass is uncertain to a factor of two @xcite .
detailed dynamical techniques are in stark contrast to estimates of dynamical mass which simply take a single velocity dispersion of the galaxy ( e.g. from a single fibre observation ) , ( e.g. * ? ? ? * ) .
it is relatively easy to extract a rotation curve @xmath10 from the observed data if the kinematics are fairly well ordered @xcite .
this can then be used to provide dynamical information about the galaxy , particularly if baryonic information is brought to bear .
however if the kinematic axes are misaligned with the photometric axes , this is often a signature of streaming motions due to a bar or an oval distortion .
a dynamical mass can still be derived , although the increased number of free parameters makes this more uncertain @xcite .
any deviations from rotational symmetry are important in their own right .
it is often very difficult to see the presence of bars , particularly in highly - inclined disc systems .
but bars are often betrayed by the inner twists of the isovelocity contours .
warps are more easily detected on large scales and almost always in hi kinematics @xcite ; however , the same effects can now be seen in deep observations of the diffuse ionized gas in the outer disc @xcite .
the physical cause of kinematic distortions can be examined by large ifu surveys which probe a variety of galaxy parameters .
for example , are distortions more likely in high density regions due to dynamical interaction .
integral field spectroscopy also enables studies of stellar kinematics that describe the observed projected stellar angular momentum per unit mass of galaxies , not possible with other techniques @xcite .
this measurement enables the separation of early - type galaxies into fast and slow rotators which are thought to have very different formation paths . a great success of the sauron and atlas-3d projects @xcite has been the discovery that most early type galaxies have significant rotation , with only @xmath11 percent ( predominantly at high mass ) being slow rotators .
@xcite have used atlas-3d to demonstrate a kinematic morphology - density relation , which shows a smooth transition of spirals to early type fast rotators with increased density , and massive slow rotators only inhabiting the highest density regions .
sami would allow such studies to be expanded to probe greater dynamic range in environment and examine a such relations as a function of stellar mass .
the rate of dark matter halo merging can be accurately estimated from simulations @xcite .
kinematic information from integral field spectroscopy can be used to differentiate between quiescent galaxies and those undergoing a merger ( e.g. * ? ? ?
* ) , using procedures such as kinemetry @xcite .
until now , this type of analysis has largely been limited to high redshift ( where the merger rate is expected to be higher ) , but with samples of @xmath12 or more galaxies , statistically meaningful estimates of merger rates can be made at low redshift .
@xcite predict that the halo merger rate at @xmath13 should be @xmath14halo@xmath15gyr@xmath15 for major mergers
( with mass ratios @xmath16 ) .
integral field observations provide a complementary approach to studies which focus on the number of close pairs to estimate merger rates ( e.g. * ? ? ?
* ) , as they probe very different phases of the merger process .
it is also be possible to look for weaker dynamical disturbances in discs due to repeated minor mergers / interactions @xcite . in this case
, it is very advantageous to extend the study to large radius as the effect of tidal perturbations scales as @xmath17 . in this case
the challenge is to achieve sufficient sensitivity at large radius ( e.g. @xmath18 scale lengths ) .
spin angular momentum from galaxy kinematics can directly probe the formation of the large scale structure of the universe and galaxy formation .
the spin of galaxy discs provides an approximation of the spin of the galaxy s dark matter halo @xcite , which is coupled to the large scale structure .
early in a dark matter halo s life , it experiences torques from the surrounding density landscape .
the spins of dark matter haloes today retain some memory of that landscape , so spin is intrinsically linked with the large scale structure .
this link may be examined in n - body simulations and observations by measuring the distribution of inferred dark matter halo spin magnitude @xcite and by the relative orientation of galaxy spin directions with each other and with the large scale structure .
n - body simulations do not predict a strong alignment between the spins of neighbouring haloes , although an alignment between galaxies has been detected in the tully catalogue of 12,122 nearby spirals @xcite .
simulations and theory predict an alignment between halo spin and the tidal field @xcite , and an alignment with features in the tidal field like filaments @xcite , sheets @xcite and voids @xcite .
any kind of alignment is predicted to be very weak , however , so could only be seen in large scale galaxy surveys .
there have been detections of spin alignments in the large scale structure reconstructed from imaging surveys @xcite , using inferred galaxy spins from disc shapes .
a tentative detection of spin alignment with filaments was found using the inferred spin of only 201 galaxies around voids @xcite and 70 galaxies in filaments @xcite , picked from the large scale structure of sdss .
discrepancies between the results found from dark matter simulations and observations indicate a difference in the way that galaxies and dark matter haloes obtain and keep their spin .
a large survey of direct spin measurements could reveal whether galaxies exhibit the same spin behaviour as dark matter haloes , and show how galaxy spin is linked to the large scale structure .
much recent observational and theoretical work has focussed on how blue galaxies can have their star formation quenched , moving them onto the red sequence .
red sequence galaxies are preferentially found in denser environments ( e.g. * ? ? ?
* ) , and star formation is also clearly suppressed at high density ( e.g. * ? ? ?
this immediately suggests environmental factors play an important role .
when a galaxy falls into a cluster , the _ ram pressure _ from the dense intergalactic medium @xcite may expel the gas from the disc , removing the fuel required for further star formation .
there are several observed examples of this in rich clusters ( e.g. * ? ? ?
* ; * ? ? ?
* ) . in moderately dense regions , such as galaxy groups , it may be that ram pressure will leave the disc intact , but can still remove gas from the halo of the galaxy .
the halo provides a reservoir of gas which can replenish the disc . without the halo gas ,
the star formation will decline and then cease as the disc gas is consumed , leading to a transition to the red sequence , in a process known as _ strangulation _
simulations suggest that this process can be efficient at removing halo gas , even in small and/or compact groups @xcite , but there is little direct experimental evidence that this is the case .
indirect evidence does point to some pre - processing of galaxies in groups before they fall into clusters @xcite , but the physical process driving this is not clear .
direct galaxy - galaxy interactions are also expected to play a critical role , with major galaxy mergers triggering star formation ( e.g. * ? ? ?
* ) and transforming the morphology of galaxies , although the fraction of galaxies undergoing major mergers ( i.e. those with mass ratios of 3:1 or less ) in the local universe is small ( e.g. * ? ? ?
on the other hand , dwarf star - forming galaxies in the local universe are often found interacting with low - luminosity objects or diffuse h i clouds , .
this appears to be the triggering mechanism of their intense star - formation activity , although only detailed multi - wavelength observations are able to reveal these processes however , environment is only one factor .
feedback from star formation and accretion onto super - massive black holes provides an internal mechanism for transformation .
this feedback provides a solution to the mismatch of the theoretical dark matter halo mass function and the observed stellar mass function ( e.g. * ? ? ?
* ) by heating and/or expelling gas in both low mass ( via star formation ) and high mass ( via black hole accretion ) haloes @xcite .
extreme outbursts of star formation or black hole accretion may be triggered by mergers or interactions ( e.g. * ? ? ?
* ) , making a link between internal and environmental effects .
once the burst is over , another mechanism is needed to suppress continued star formation .
the best suggestion for this is mechanical feedback from jets emitted by super - massive black holes ( e.g. * ? ? ?
* ) , but this only appears to be efficient in massive galaxies . as well as these active processes , the environment has an indirect influence via formation age .
galaxies in high density regions form earlier and so have had more time to evolve ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in the absence of other effects
, we would then expect to see galaxies in high density regions having older stellar populations . disentangling these varied influences on galaxy formation
is far from trivial .
however , studies of the spatial distribution of instantaneous star formation rates , integrated star formation ( via stellar population ages ) and metallicity ( both gas and stellar ) provide considerable insight . importantly , ram - pressure removal of gas implies that the truncation of star formation is an _ outside - in _ process .
gas is preferentially removed in the outer parts of galaxies , which are less gravitationally bound .
this may be a short - lived feature of the galaxies in dense environments if the gas is eventually completely removed .
alternatively , stripping can occur over the lifetime of a galaxy if the gas is puffed up by the internal star formation ; in this case , even a rarified external medium can remove gas from the galaxy @xcite .
globally , the expectation would be that galaxies form inside - out , and this implies age and metallicity gradients which are observed .
one approach that has been explored in some detail as an alternative to spatially - resolved spectroscopy is the so - called ` pixel - z ' technique @xcite .
this approach , analogous to photometric redshifts ( ` photo - z ' ) , uses a library of template spectral energy distributions ( seds ) to fit the observed optical and near - infrared colours of individual pixels within resolved galaxy images .
this technique has been used with some success to explore the environmental dependence of star formation in galaxies .
@xcite find that , globally , the suppression of star formation in high density environments ( e.g. , * ? ? ?
* ; * ? ? ?
* ) seems to occur primarily in the most strongly star - forming population , and to be evidenced by a suppression in the _ inner _ regions of galaxies .
@xcite demonstrate that this effect seems to hold independently for both early- and late - type galaxy populations , and that the suppression in star formation can not be explained solely by the well - known density - morphology relation ( e.g. , * ? ? ?
there are significant limitations , however , to the ` pixel - z ' approach .
these are related to implicit assumptions made by the technique , such as each pixel being represented by an isolated single stellar population with a simple exponential star - formation history @xcite . as a result ,
the method can not effectively measure the instantaneous star formation rate , which can be traced spectroscopically ( e.g. by emission ) .
ifu spectroscopy allows us access to both current star formation ( via emission lines ) and integrated star formation history ( via stellar age and metallicity ) .
examining the radial dependence of the mean stellar age and metallicity gradients tells us when and where the stars formed in these galaxies , along with a fossil record of the galaxy merger history .
the mean stellar age is effectively a luminosity - weighted integral of the star formation history whilst the stellar metallicity gradient provides an indication of its merging history ( e.g. * ? ? ?
if environmental effects are responsible for the cessation of star formation , then we would expect red sequence galaxies to have younger central ages with past major mergers sign - posted by shallow negative metallicity gradients ( i.e. lower metallicities in the outskirts ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
a full picture of the physical processes involved in the fuelling of accretion onto super - massive black holes , resulting in the phenomenon of an active galactic nucleus ( agn ) , is still elusive .
it is now known that most galaxies contain super - massive black holes , with typical masses a million to a billion times that of the sun ( e.g. * ? ? ?
* ; * ? ? ?
the mass of the black hole correlates well with the mass ( or velocity dispersion ) of the bulge or spheroidal component @xcite in a galaxy .
this implies an intimate connection between the build up of stellar mass in galaxies and their super - massive black holes .
the nature of the connection between star formation and agn has long been debated ( e.g. * ? ? ?
* ) , although a resolution remains elusive .
the most luminous agn ( i.e. bright quasars ) require @xmath1910@xmath20 solar masses of gas deposited in their central regions on time - scales of @xmath410@xmath2110@xmath22 years ( e.g. * ? ? ?
* ) , requiring major galaxy - wide perturbations @xcite , and no doubt triggering substantial star formation .
in contrast , low - luminosity agn require relatively small amounts of gas , which can be supplied by internal stochastic processes , such as the accretion of cold molecular clouds @xcite .
the agn population evolves strongly with cosmic time .
this is most clearly seen in evolution of luminous quasars from @xmath23 to @xmath24 @xcite , which have a strong peak in space density at @xmath25 .
lower luminosity agn show less severe evolution , and peak in space density at lower redshift .
this downsizing appears qualitatively similar to that seen in the formation of galaxies @xcite . in the local universe , accretion onto black holes
is dominated by systems with low black hole masses , @xmath26m@xmath27 , but which are typically accreting within an order of magnitude of the eddington limit @xcite .
in contrast , higher mass black holes in the local universe are accreting with a characteristic timescale substantially longer than a hubble time .
luminous agn ( as measured by their @xmath75007 luminosity ) are found to have younger stellar populations @xcite , as measured by their @xmath28(4000 ) and h@xmath29 line indices from sdss spectra .
it is not clear whether this younger stellar population is centrally concentrated , or is distributed more widely across the host galaxy , as the sdss fibres subtend a physical scale of @xmath30 kpc at @xmath31 .
ifs data has the ability to explicitly examine the distribution of star formation and stellar population ages across galaxies and investigate whether these are related to accretion rate .
lower luminosity agn tend to have spectra typical of low - ionization nuclear emission - line regions .
while there is evidence that these form a continuous sequence with seyferts @xcite , there substantial evidence that liner like emission is extended and may be powered by ionization from agb stars @xcite . a large galaxy survey with spatially
resolved spectroscopy could resolve this issue .
if liner emission is not due to a central agn in most cases this would require substantial re - interpretation of recent work on low redshift agn .
at an even more fundamental level , the fraction of galaxies which host agn is only well defined locally .
approximately @xmath32 per cent of these very local galaxies host agn . at larger distances contamination from off nuclear emission increasingly reduces the sensitivity to weak nuclear emission lines .
a large ifs survey would enable apertures of fixed metric size to be defined , enabling robust agn rates as a function of redshift to be determined . while there is a good theoretical basis for galaxy mergers triggering luminous agn ( e.g. * ? ? ?
* ) , the evidence for this is mixed .
an alternative pathway is via violent disc instability in self gravitating discs ( e.g. * ? ? ?
* ; * ? ? ?
disc instability is at least likely to play a role at high redshift where cold streams can dominate the mass accretion onto discs @xcite .
indeed , a large fraction of high redshift star forming galaxies appear to be discs ( e.g. * ? ? ?
* ; * ? ? ?
* ) , rather than mergers . in low redshift samples agn activity
is not enhanced by the presence of a nearby companion @xcite , while star formation is @xcite .
this is a somewhat surprising result , and may point to a difference in timescale between the onset of star formation and the agn , with the agn occurring later , after the merger has taken place .
indeed , ifs observations of local galaxies with observed outflows demonstrates that the agn timescale is significantly longer than the starburst timescale @xcite .
large scale ifs observations can directly address the issue of agn fuelling by examining the kinematic properties of agn hosts , and searching for evidence of disc instability and/or merging . in this regard
it will be particularly important to span a range of accretion luminosities in order to examine whether there is a change from secular evolution at low accretion rate to mergers at high accretion rate .
the accretion of gas onto galaxies remains a largely unsolved problem . whether gas enters the galaxy potential in a hot phase and
then cools down @xcite , as a warm rain @xcite , or as an hi complex like the high velocity clouds in the galactic halo , or all of the above , is unclear . after a review of the evidence , concluded that the outer warps of hi discs were some of the best evidence of ongoing disc accretion .
once the gas settles into the galaxy , the outstanding issues are how the gas feeds into the nuclear regions and , in particular , onto a central black hole . in a survey of @xmath33 galaxies , sami offers the prospect of studying nuclear activity ( agn , starburst , liner ) and star formation within the context of the extended galaxy .
the kinematic signatures of outer warps and inner bar streaming are relatively easy to pick out in hi @xcite or in ionized gas @xcite .
thus , we can now directly associate this activity with large - scale disc disturbances , assuming these exist .
traditionally , in large galaxy surveys , the association of activity and dynamical disturbances is made from the proximity of galaxies in position and redshift space ( e.g. * ? ? ?
* ; * ? ? ?
. with full kinematic information a much more direct determination of the triggering of activity will be possible .
an alternative approach is to study the impact of the inner disc on the extended properties of galaxies ( e.g. * ? ? ?
* ; * ? ? ?
jets carry energy , and winds carry gas and metals , far from the nucleus . in a recent integral field study of ten agn and starburst galaxies , @xcite find that starburst winds are largely shock ionized , while agn winds show the hallmark of photo - ionization by the accretion disc , clearly indicating that the starburst phenomenon is very short . for one of the objects observed on the sami commissioning run , we see the ionization characteristics typical of nuclear activity for gas off the plane of an inclined disc ( see section [ sec : gals ] and fogarty et al .
, in prep . ) .
a large - scale wind is confirmed by the broad emission - line profiles along the minor axis .
this is a remarkable testament to the power of spatially - resolved kinematic and ionization information . in the full survey ,
sami is likely to uncover hundreds of new outflow sources connected either to nuclear activity or inner disc star formation .
there is an even rarer class of galaxies with disc - wide winds that sami will also inevitably add to @xcite .
a number of edge - on spiral disc galaxies have vertically extended ionized gas in their haloes @xcite .
the reynolds layer in our galaxy , recently mapped by the wham survey telescope @xcite , is a good example of this phenomenon ( see also * ? ? ?
the ionization characteristics of this gas does not appear to be consistent with any known mechanism ( i.e. uv photo - ionization by hot young stars , radiation from old supernova bubbles , shocks from supernovae , cosmic ray heating , or radiation pressure on dust grains in the disc ) .
the gas may arise from some kind of disc - wide interaction between the disc and the hot halo , presumably driven by processes related to star formation in the disc .
but there is also the prospect that some of this gas is related to warm gas accretion onto the disc @xcite or involved in a large - scale circulation or recycling of gas through the halo @xcite . as with the study of galaxy winds , a large sami survey has the potential to greatly increase the sample of known galaxies with vertically - extended warm discs . with a larger sample
, it will be possible to correlate the presence of these discs with the disc star formation rate , nuclear activity and galaxy mass .
historically , telescopes were used to observe one source at a time . but with technical advances in optical fibres , it was realized in the early 1980s that many sources could be observed simultaneously across the telescope focal plane by precisely positioning fibres in the field @xcite .
this led to an explosion in wide - field spectroscopic surveys , notably including the 2-degree field galaxy redshift survey ( 2dfgrs ; * ? ? ? * ) , 2-degree field qso survey ( 2qz ; * ? ? ? * ) , 6-degree field galaxy survey ( 6dfgs ; * ? ? ? * ) and sloan digital sky survey ( sdss ; * ? ? ? * ) amongst several others . between them ,
such surveys have obtained spectra for approximately 1.5 million extragalactic targets .
new instruments recently commissioned ( e.g. lamost ; * ? ? ?
* ) or in construction ( e.g. virus ; * ? ? ?
* ) are able to observe thousands of sources at a time .
considerable advances have been made possible by the 2dfgrs and sdss , which use a single optical fibre per galaxy .
however , with fibre diameters of 2 and 3 arcseconds respectively , these projects sample less than half the light from a galaxy at the median distance of the surveys .
these single apertures limit the surveys in two ways .
first , it is impossible for single - fibre surveys to measure spatially varying spectral properties , which prohibits the study of crucial observables such as kinematic merger rates , galaxy rotation and dynamical mass , star formation gradients , metallicity gradients , age gradients and detection of galaxy winds and/or outflows .
the second limitation is that with single - fibre spectroscopy the measured signal depends on many things : ( i ) intrinsic properties , like source luminosity , size and distance ; ( ii ) atmospheric conditions , particularly seeing ; ( iii ) instrumental properties , like fibre aperture size and positioning accuracy , and optical focus over the field ; ( iv ) telescope properties , such as pointing and guiding precision ; and perhaps other effects .
many published papers make the mistake of assuming the surveys provide _ spectrobolometry _
( i.e. the spectrum of the total light output by the source ) rather than the spectrum from an ( often ill - defined ) spatial sample of the source .
the inherent dangers of aperture effects have long been known in astronomy , but have often been under - appreciated or ignored .
* see their fig .
8) clearly demonstrate that the single - aperture fibre spectra from a typical galaxy survey may be only weakly correlated with the photometric classification .
aperture biases can manifest themselves in two ways .
the fibres subtend an increasing linear size with increasing distance of the galaxy , potentially causing spurious evolutionary effects ( and spurious luminosity - dependent effects in a flux - limited sample ) .
second , important galaxy properties , such as star formation rate and metallicity can have substantial gradients , meaning that observations of just the central regions are not representative of global values ( e.g. * ? ? ? * ) .
integral field spectroscopy allows us to gather data on key observables that are simply inaccessible to single apertures .
ifus are usually a single monolithic array of lenslets , with the light fed to a spectrograph via optical fibres ( or alternatively an optical image slicer ) . as a consequence
, they can only target one object at a time and so ifu surveys have typically only targeted a few tens of galaxies in specific classes ( e.g. * ? ? ?
* ; * ? ? ?
the process of galaxy formation and evolution is inherently complex , with the observed galaxy properties depending on a large number of parameters , such as host halo mass , stellar mass , merger history etc . as well as the multi - dimensional nature of galaxy properties ,
there is inherent stochasticity in the process .
this is at least in part due to our inability to accurately trace the formation history of individual objects , but also derives from inherently non - linear physics such as that involved in the collapse of molecular clouds to form stars .
the multi - dimensional nature of the galaxy population , combined with this stochasticity ( which at some level could be considered as extra hidden parameters ) , means that large surveys are required to extract the key relations between physical properties .
the success of surveys such as 2dfgrs and sdss has in large part been due to their ability to ` slice and dice ' the galaxy distribution and still have statistically meaningful samples in each bin of the parameter space .
the need to extend integral field observations to large samples is well understood in the above context , and has driven recent projects such as atlas-3d @xcite and califa @xcite .
these impressive projects are limited by their use of monolithic integral field units , which means that targeting more than a few hundred objects is prohibitively expensive in terms of telescope time .
this limitation naturally drives us to multi - object integral field spectroscopy ( i.e. multiplexed ifus ) , the subject of this paper . a key challenge for multi - object integral field spectroscopy is to obtain sufficient signal - to - noise at low surface brightness levels for all the targets observed simultaneously in a given pointing . in this section
we present some preliminary investigations of the properties of potential sami targets .
we will use the recent sersic fits and bulge - disc decomposition carried out by @xcite on the sdss photometry .
first we consider a simple apparent - magnitude - limited sample with @xmath34 ( extinction corrected , where @xmath35 is the galaxy sdss @xmath36-band magnitude derived from a sersic model fit to the photometry ) .
this limit was chosen as the galaxy surface density approximately matches the density of ifus in sami .
the distribution of half - light radius , @xmath37 , and surface brightness at @xmath37 is shown in figure [ fig : size_sb ] .
for such a sample the sami hexabundle ifus reach to @xmath38 for all but the largest 10 percent of galaxies ( blue dashed line ) , while @xmath38 is sampled by at least 3 ifu elements for all but the smallest 10 percent of galaxies .
in other words , for the central 80 percent of this sample , sami can give spatially resolved spectroscopy out to at least @xmath38 .
the median surface brightness at @xmath38 is @xmath39magarcsec@xmath40 .
( panels a and b ) and for a volume - limited sample with @xmath41 and @xmath42 ( panels c and d ) .
the black solid line is the median and the black dotted lines are the 10th and 90th percentiles .
the blue solid line is the radius of a single fibre core in sami and the blue dashed line is the radius of a 61-core hexabundle .
[ fig : size_sb_z],title="fig:",width=264 ] ( panels a and b ) and for a volume - limited sample with @xmath41 and @xmath42 ( panels c and d ) . the black solid line is the median and the black dotted lines are the 10th and 90th percentiles .
the blue solid line is the radius of a single fibre core in sami and the blue dashed line is the radius of a 61-core hexabundle .
[ fig : size_sb_z],title="fig:",width=264 ] we show how size and surface brightness vary with redshift for this sample in figures [ fig : size_sb_z]a and b. the typical sizes in arcseconds of the galaxies stay relatively constant with redshift , largely because the @xmath36-band limit selects more massive ( and therefore larger ) galaxies at higher redshifts .
a natural alternative is to select a volume - limited sample , which is shown in figures [ fig : size_sb_z]c and d. in this case we choose @xmath41 and @xmath42 , which gives similar numbers of targets ( i.e. similar surface density ) to the @xmath36-band cut used above . in this case
we see , unsurprisingly , that the galaxies are smaller at high redshift , but that the median @xmath37 is more than twice the radius of a fibre core ( and more than three fibre cores out to @xmath43 ) .
an apparent - magnitude - limited sample can be considered as a set of volume - limited samples with small redshift intervals and absolute magnitude limits that vary with redshift .
it is highly likely that the optimal solution for targeting galaxies for a multi - object ifu instrument involves taking multiple volume - limited samples , allowing optimal use of the ifu fov while broadly populating the distribution of galaxy stellar mass .
one issue that needs to be considered is that in an apparent - magnitude - limited sample , objects inevitably pile up around @xmath44 , so suitable sampling may need to be implemented to efficiently cover a wide range in stellar mass .
last considerations in target selection include whether to sample all the galaxies in a given volume or to choose field locations which uniformly sample the distribution of galaxy environment .
the challenge with the latter approach is that there are a variety of different local density estimators and the relation between them is not trivial ( see brough et al .
, in preparation ) .
there is also a challenge to balance the requirements of maintaining sufficient spatial resolution below @xmath38 , as well as reaching @xmath45 which is a requirement for reaching the turn - over in the rotation curves of disk galaxies ( e.g. for tully - fisher analysis ) .
separate samples within the same survey area that can be observed concurrently may be the most natural solution to this problem
. further discussion of the detailed sample selection for a large sami galaxy survey will be deferred to a future paper .
with a view to providing the first on - telescope demonstration of hexabundle technology , the australian astronomical observatory ( aao ) and the university of sydney have collaborated on the development of the sami instrument for the 3.9 m anglo - australian telescope ( aat ) .
sami uses 13@xmath4661-core hexabundles that are mounted on a plug - plate at the 1-degree fov triplet corrector top - end focus of the aat . at f/3.4 , with 105@xmath47 m core diameter fibres , each hexabundle samples a 14.9arcsec diameter field at 1.6arcsec per fibre core . at the output end , a total of 13 v - groove slit blocks are mounted at the slit of the aat s aaomega spectrograph ( @xcite ; also see section [ sec : aaomega ] ) .
each slit block includes 63 fibres ( all the fibres from one hexabundle plus two fibres for sky subtraction ) .
a fusion - spliced ribbonised fibre cable of length @xmath442 m joins the two instrument ends together .
a near real - time data pipeline , based on the 2dfdr code @xcite , has been written to reduce the data .
the following subsections describe the instrument requirements , fibre cable , the hexabundles , the prime focus unit , the sami field plates , the aaomega spectrograph , the instrument control and data reduction software .
the sami instrument was designed to be a technology demonstrator and to carry out significant science programmes . as a result
, the final instrument design is influenced by a mix of scientific and technical constraints .
a key constraint was to develop the system on a rapid time - scale , which naturally led to the use of the plug - plate system and the already available aaomega spectrograph .
the flexibility of aaomega , with a range of resolutions and wavelength settings also enables a variety of science . in order to make a substantial advance over previous facilities , the multiplex of the system had to be at least an order of magnitude better than previous monolithic ifus ( i.e. sami required at least @xmath48 ifus ) .
the median optical seeing at the aat is @xmath49 arcsec , so the fibre cores were approximately matched to this ( 1.6 arcsec diameter ) .
although undersampling the seeing , this is generally preferable to having smaller core sizes for a number of reasons : i ) larger fibre cores provide more independent resolution elements , ii ) smaller fibre cores will lead to data being read - noise limited in the blue , iii ) larger fibre cores provide better surface brightness sensitivity , iv ) larger fibre cores allow higher fill factors , given a fixed minimum fibre cladding ( i.e. @xmath50 m in our case ) , v ) critical sampling of the seeing can still be achieved with dithered exposures .
given the above , the final key instrument design decision was the number of fibres per ifu .
the choice of 61 fibres per ifu was made largely on the basis of the known capability to manufacture such bundles .
however , this number of fibres also matches other requirements .
larger numbers of fibres per bundle would have restricted the multiplex given the fixed aaomega slit length , and the chosen 15 arcsec ifu diameter provides a good match to the scale length of galaxies in local samples chosen to match the surface density of sami ifus within the 1degree diameter field of view ( see section [ sec : size ] ) .
if designed for a single experiment , one option would have been to have a variety of bundle sizes to match the specific galaxy size distribution of the target sample . however , for ease of manufacture , and to maintain multiplex and flexibility , we chose to have all the ifus the same size .
sami is mounted at prime focus on the aat and feeds the aaomega spectrograph located in the coude room , requiring a fibre cable run of @xmath442 m . within the fibre
bundle a total of 819 fibres are used : 793 for the hexabundles ( 13@xmath4661 fibres ) , and 26 for the sky fibres .
the hexabundles were each supplied with a fibre pigtail length of 1 m encased in a reinforced furcation tube .
each of the hexabundle and sky fibres were fusion spliced onto a fibre of length @xmath442 m that is terminated in a v - groove block at the spectrograph entrance slit ( see section [ sec : aaomega ] below ) .
for this fibre run we chose a fibre ribbon cable to minimise the required assembly effort .
each 250micron thick ribbon contains 8 fibres , and 8 ribbons are used for each of the 13 units ( one hexabundle plus two sky fibres ) . to minimize losses in the splicing process , and because of its ready availability , we matched the ribbonised fibre type to that used in the hexabundles ( thorlabs afs105/125y ) .
a ` splice - box ' is mounted on the internal wall of the telescope top - end barrel that incorporates 13 closed - cell polyethylene trays to secure and protect the individual splices of each unit .
the inputs to this box are the 26 sky fibres ( each individually sleeved with reinforced pvc furcation tubing and terminated with sma connectors ) and the 13 individual hexabundle tubes .
the output of this box is the fibre bundle .
the bundle is protected during its run ( which goes through the top end telescope ring , beside the primary mirror support , and through the declination drive axis to the coude room ) by an inner covering of braided cable sleeving and an outer double - split nylon conduit giving light - weight yet relatively strong protection .
though the use of ribbonised fibre worked well in terms of handling and assembly , tests showed that ribbonising caused a loss due to focal ratio degradation ( frd ) , as discussed in section [ sec : frd ] .
the fibre type used provides relatively good throughput above 450 nm , though it suffers higher absorption losses in the uv than other fibres more commonly used in astronomical instruments .
the current fibre cable described in this paper will be replaced in the first half of 2012 to provide much improved blue throughput .
m. surrounding the fibre cores is a glass ferrule ( black ring ) , which in turn is surrounded by the central steel pin of the sma connection , which extends beyond the edge of the image .
[ hexanum],width=302 ] @xcite introduced a new imaging fibre bundle optimized for low - light astronomical applications .
the hexabundle incorporates several technological innovations in order to achieve a high fill factor .
a primary goal of the hexabundle design is that the performance of individual multimode fibres should be as good as the same fibre in isolation .
consequently , we were forced to reject an earlier design where the fibre cores were forced into non - circular shapes because the added focal ratio degradation ( numerical aperture up - conversion ) was found to lead to significant light loss @xcite . within each hexabundle
, the 61 circular fibres ( with numerical aperture , na=0.22 ) are lightly fused together and infilled with low - stress glue .
the multimode fibres have relatively small core diameters of 105@xmath47 m . in order to achieve a high filling factor ( 75 percent )
, it was necessary to reduce the cladding to only 5@xmath47 m . the length of the fused region is very short ( @xmath5130 mm ) to minimize cross - talk between the fibre cores .
the fibre bundle is held within a reinforced flexible plastic tubing that is both strong and light - weight .
the optical head is supported by a stress - relieving sleeve and is inserted into an sma connector .
this is to allow the fibre bundle to be manually attached to the field plate with relative ease .
the repeatable positioning accuracy made possible by the sma connector is much better than a fibre core size .
the use of bare fibre hexabundles directly at the telescope focal plane is the most novel element of our instrument design .
there are several advantages to such an arrangement as an alternative to fibre - lenslet coupled ifus that have been used elsewhere ( e.g. * ? ? ?
first , the hexabundle is fabricated as a one - step process , whereas a fibre - lenslet system requires the fibre bundle to be manufactured and then bonded to the lenslet array(s ) . secondly , there are no optical elements required in our hexabundle design .
lenslet or microlens array ifus require at least 2 surfaces and more often up to 8 surfaces before the fibre face , to adjust for plate scale and to preserve telecentricity of the telescope beam into the fibre .
these extra surfaces add to system loss .
a trade - off must be made against the potential for extra losses in the hexabundle ( e.g. from frd - see below ) and from its lower fill factor .
for sami , as we are feeding the hexabundles with a fast beam the frd should be minimised ( if used with the appropriate fibre type ) . finally , the hexabundle solution offers the opportunity of reduced pitch between ifus relative to a lenslet solution that must work at a higher magnification .
lcccc fibre & & + & red & blue & red & blue + & % & % & 90% ee & 90% ee + hexabundle 15 : & & & & + core 1 alone & @xmath52 & @xmath53 & 0.003 & 0.009 + core 1 plus ribbon & @xmath54 & @xmath55 & 0.028 & 0.037 + core 1 plus ribbon & & & & + after run & @xmath56 & @xmath57 & 0.067 & 0.073 + + core 18 plus ribbon & @xmath58 & @xmath59 & 0.026 & 0.034 + core 18 plus ribbon & & & & + after run & @xmath60 & @xmath61 & 0.076 & 0.082 + + core 6 plus ribbon & @xmath62 & @xmath63 & 0.028 & 0.037 + core 6 plus ribbon & & & & + after run & @xmath64 & @xmath65 & 0.037 & 0.044 + predicted performance & 84 & 62 + with replacement fibre & & + 42 m bare fibre & @xmath66 & @xmath67 & 0.008 & 0.018 + 42 m ribbonised & @xmath68 & @xmath69 & 0.022 & 0.031 + focal ratio degradation ( frd ) increases the size of the light cone coming out of a fibre compared to that put in .
the worse the frd , the more light will be lost from the f/3.15 acceptance cone of the aaomega spectrograph .
however , frd can be partially controlled by minimising the stresses on the fibres when installing them in the system .
an additional loss comes from the throughput of the fibres , which is primarily dependent on the fibre type and length . in this section
we determine the performance of the fibre cable and hexabundles by analysing frd and throughput .
as will be seen below , the dominant source of losses is the fibre cable , rather than the hexabundles .
the frd and throughput were measured for a sami hexabundle using an led source that was fed through bessel @xmath70 and @xmath71 filters ( centred at approximately 435 nm and 625 nm ) and re - imaged to form an f/3.4 beam .
this was then input into several hexabundle cores in turn .
the output fibres were imaged in the far - field using an sbig camera . after flat - fielding the images ,
the centre of each output spot was fitted .
encircled energy was calculated in concentric circles about the centre position using software packages within iraf .
we initially compared the performance of a hexabundle alone to the same hexabundle when spliced to the 42 m of ribbonised cable with a slit block attached ( hexabundle 15 , ` alone ' vs ` plus ribbon ' in table [ frd ] ) .
for three different cores ( numbers 1 , 6 and 18 ; see figure [ hexanum ] ) , the throughput dropped to @xmath72 percent in the red and 41 percent in the blue with the addition of the ribbonised cable plus slit block . figure [ navsee ] and table [ frd ] show that this drop is identified with a significant increase in frd , with the na up - conversion ( at 90 percent encircled energy ) for the hexabundle with ribbonised fibre measured to be significantly worse than for the hexabundle alone .
the end finish on the bundle alone was a cleave , but the measurement through the ribbonised fibre and slit block had the advantage of a polished end finish which should improve the frd , so the frd introduced by the ribbonised fibre may be a little worse than these numbers indicate . in na ) .
the dashed curves ( with errors of @xmath73 ) are for fibres 6 , 1 and 18 ( same colours as above ) , after the ribbonised fibre was put into the cable sleeve and outer nylon conduit and had been transported to the telescope , installed and used for the first commissioning run.,width=264 ] in order to differentiate between the effect of the 42 m of fibre and the ribbonising , we separately compared bare and ribbonised fibre throughputs ( table [ frd ] , lower panel ) .
this was done by measuring the throughput from 10 m of bare fibre of the same type as used in the hexabundles and ribbonised fibre ( afs105/125y ) .
the 10 m was then cut , spliced together and measured to give the splice loss .
then the splice was cut and an additional 42 m was spliced in place .
this was then cut out and 42 m of ribbonised cable was spliced in instead .
the throughput could then be compared for the bare and ribbonised cable after accounting for the initial 10 m .
having the initial fibre in place meant that the frd results for the bare and ribbonised fibre were not affected by coupling into the fibre or end finishing effects as these remained the same for both tests .
variations in throughput due to the different splices have been taken into account in the errors .
table [ frd ] shows that while the bare fibre results will include end effects , the ribbonising of the fibres results in significantly worse frd .
the bulk of the loss in throughput is due to the length of fibre ( 78 percent throughput in red and 58 percent in blue ) , however the frd from ribbonising results in less of that light coming out within the f/3.15 acceptance cone of the spectrograph .
once the hexabundles were spliced to the ribbonised cable and slit blocks ( ` plus ribbon ' in table [ frd ] ) , the throughput was lower than that measured for the 42 m of ribbonised cable alone .
this is due to an increase in the frd above that of the ribbonising .
the small additional frd is therefore due to the hexabundle and slit block .
mechanical stress on the fibres when mounted in the slit block can increase the frd ( see for example * ? ? ?
* ) , and hence decrease the throughput within f/3.15 . ,
width=124 ] during the commissioning run it was noted from aaomega images that the throughput of the fibres on the outer edge of each ribbonised cable was lower than that of the fibres in the centre of a ribbonised cable .
therefore two fibres on the edge of a ribbon and one in the centre of a ribbon were tested after sami had been taken off the telescope .
the testing method was identical to that described above , and the three fibres tested were the among the same fibres from the bundle that was tested before the sami commissioning run . in figure [ navsee ] and table [ frd ]
it is notable that while the fibres had similar performance before commissioning ( see ` core plus ribbon ' in the table ) , after commissioning ( see ` core plus ribbon after run ' ) the frd was substantially worse for the two fibres on the edge of a ribbon ( fibres 1 and 18 ) .
meanwhile , the fibre in the centre of the ribbon ( fibre 6 ) showed only a small increase in na up - conversion compared to the previous results .
the throughput within f/3.15 for the cores on the edge of a ribbon decreased from 64 - 67 percent and 41 - 43 percent ( red and blue respectively ) before the sami installation to 37 - 44 percent and 22 - 24 percent after the sami commissioning run .
however , the hexabundle core in the centre of a ribbon ( core 6 ) had a comparatively higher throughput of 56 percent and 32 percent .
we believe that the increased frd for the edge fibres is due to stresses in the ribbonised cable when it was fed into the braided cable sleeving and outer nylon conduit that protects the fibre run .
as the fibre bundle was moved during transport , installation and commissioning , the ribbonised bands could be bent in all directions , in which case the edge fibres come under more stress than the centre
. the throughput of sami will soon be significantly improved by the replacement of the fibre run .
the loss due to the fibre length will be reduced , particularly in the blue , by using a different type of fibre .
the fibres being considered will result in up to 14 percent higher throughput at the blue end of the spectrum .
in addition , if the effects of ribbonising and frd stresses from the ribbonised cable are removed , the throughput would further increase by up to @xmath427 percent , perhaps doubling the current blue throughput .
an alternative protection will be required for the 42 m fibre run , and that may introduce some frd losses , but it is being designed to have less of an effect than the ribbonising . within the aat prime - focus top - end , the triplet corrector and the prime focus camera were refurbished for the sami instrument .
originally built for the commissioning of the aat in the early 1970s , they provide a 1-degree fov with a plate scale of 15.2arcsec / mm . with 61@xmath46105@xmath47 m cores ,
this provides a 15arcsec diameter fov for each hexabundle and a sampling of 1.6arcsec per fibre core ( see figure [ fig : triplet ] ) . with a total of 39 fibre positions in each field ( 26 sky positions and 13 object positions ) , we chose to use a plug - plate assembly rather than a robotic positioning system , as the operational overhead and down - time between fields for reconfiguration are both relatively low , particularly as we are targeting long ( @xmath42-hour ) integrations and have a connectorised fibre system . the sami plug plates are pre - drilled 3mm - thick brass discs with through - holes at each object / sky location .
each hexabundle and sky fibre is terminated in a sma screw - thread fibre connector .
a mating connector is installed in each plug - plate at each position .
two galaxy fields ( i.e. 26 objects ) are pre - drilled on the plate along with a set of 26 blank - sky locations common to both galaxy fields . for the proposed integration time of 2hours per field
, this means that 2 plates ( and 1 plate exchange ) are required each observing night .
the down - time between fields recorded during the initial commissioning run was less than 30 minutes .
further investigation found that it was possible to include 4 fields per plate , so that no plate exchange is required during the night .
the plug - plates are installed within an assembly that is kinematically mounted to the prime focus camera ( see fig . [
fig : pfu ] ) . due to the relatively large fov of each hexabundle ,
there is not a strict requirement on the positioning accuracy of each hexabundle central fibre .
however , we aim to reach a total positional accuracy of half the core diameter of a single fibre ( i.e. 0.8arcsec ) . the accuracy achieved is determined by several factors that include : the hexabundle concentricity ; the connector concentricity ; the plug - plate machining accuracy ; and the plug - plate thermal expansion .
these factors are all controllable to within much less than a fibre core .
additional position errors are introduced via i ) a rotation offset , which is corrected via a fine - thread micrometer rotation adjustment between the plug - plate assembly and the prime focus camera , ii ) an asymmetric radial error arising from an @xmath74@xmath75 offset of the plug plate from the telescope optical axis , for which the tolerances are quite large and iii ) a symmetric radial error arising from an incorrect plate scale , which can be corrected after measurement by producing a new distortion map . for image reconstruction
it is important that the rotational alignment of each hexabundle is known .
because the sma connector has no rotation adjustment capability and is not rotationally keyed , the orientation of the hexabundles was determined by eye as they were inserted into the plug plate .
laboratory and on - sky tests demonstrated that an accuracy of less than half a core in the outer hexabundle ring ( @xmath7610degrees ) was possible with this technique .
the sma connectors will shortly be replaced by keyed fc connectors to eliminate this limitation . for acquisition and guiding
, we use a ccd camera mounted on a gantry above the plug - plate that views ( via an optical relay ) the central region of the field through a hole in the plate .
this camera ( an 800@xmath46600 pixel watec # 120n ) provides a video output signal that can be integrated from frame rate up to @xmath410 seconds and is compatible with the existing aat control system guiding software .
it reaches @xmath414th magnitude over a sky fov of 150arcsec diameter ( providing a sufficiently high sky coverage factor for most galactic latitudes ) and has a sampling of 0.3arcsec / pixel . using the manual plug - plate method for positioning the hexabundles in sami presents a number of challenges for the field allocation and plate manufacturing .
here we describe the sami field configuration methodology .
the plates are manufactured using brass of thickness 3 mm and approximate diameter 240 mm ( corresponding to the 1 degree fov ) .
the fibre connectors have an assigned footprint of 15 mm to allow access for installation and removal .
the instrument uses two distinct physical plate types : science and calibration .
each plate is configured with multiple stacked fields for efficiency gains .
there is a central 10 mm hole in each plate for the guide camera to image a sky region of diameter @xmath4150arcsec .
the science plates consist of two stacked galaxy fields each with 13 galaxy targets for the ifus and about 5 field alignment stars , also targeted with ifus .
the field alignment stars did not share the same field centre as the galaxy targets . during commissioning ,
when the plate was first observed , ifus were positioned at the location of the alignment stars in order to check for consistent rotation and plate scale between each plate . once confirmed , the ifus were re - allocated to galaxy targets and the telescope moved to the science field centre .
each plate also contained 26 sky fibre positions shared between the two galaxy fields .
the science plate is configured for consecutive fields requiring only the re - positioning of the 13 ifus onto new galaxy targets .
each science plate then has a total of @xmath77 drilled holes for the installation of the sma fibre connectors ( with a minimum separation of 15 mm ) .
the astrometric calibration plates were constructed in a similar way , but in that case each plate contained four overlapping fields and four visual alignment fields .
the process for generating a single field was as follows : + ( i ) for each ra region , convert the ra and dec .
coordinates to angular distance coordinates and calculate the pairwise distance matrix of all targets .
+ ( ii ) iterate over each target using the distance matrix to extract targets within a 0.5degree radius and to remove targets with separations @xmath7615 mm ( @xmath4228arcsec ) .
+ ( iii ) count the number of targets and identify this as a candidate field if the total is equal to or greater than @xmath78 , where @xmath78 is the number of sources required ; @xmath78=14 for astrometric calibration fields ( 13 ifu targets plus central guide star ) , @xmath78=3 for visual alignment fields ( two visual targets and a central guide star ) and @xmath78=13 for science fields .
+ ( iv ) for a candidate field , save the data and assign the field to that ra region ; eliminate the field s allocated objects from the list of potential targets ( in order to produce fields with unique sets of targets ) .
once individual fields were defined , multiple fields were stacked on a single plate ( e.g. for science fields the plates contain @xmath79 galaxies plus @xmath80 alignment fields ; for astrometric calibration fields the plates contain @xmath81 calibration targets plus @xmath82 visual alignment fields ) .
fields were grouped into ra regions ( typically of size 20degrees in ra ) based on observation times .
the ra regions were ordered based on the best observing time for each region ( i.e. the time of lowest airmass ) .
then fields were tested in turn to see that the allocated targets did not overlap , so that they could be stacked onto an individual plate .
the priorities of the targets in each science field are summed to give a figure of merit for all possible plate permutations .
the permutations can be filtered for field duplicates and sorted by merit . once the plate configuration was defined , we determined the 26 sky fibre positions that are shared for each field .
this was based on a grid of regularly spaced angular positions relative to the plate centre ( e.g.@xmath83 grid points over the field ) , which were then converted to ra and dec
. given each field centre , and to relative plate coordinates in microns ( the same for each field ) .
sky fibre grid locations were eliminated if they overlapped with the science targets in plate coordinates .
supercosmos images @xcite were examined at the ra and dec . of each remaining sky grid position for each field and if no source was found within a 24arcsec window then the sky position was accepted . from these remaining candidates , 26 sky fibre locations
were then selected for targeting . for each stacked field
, we then determined suitable guide stars ( @xmath84@xmath7614 ) in the central 150arcsec fov using the aladin sky atlas ( http://aladin.u-strasbg.fr/ ) to search the usno - b1 catalogue and examine dss images .
after the above process , we take the entire set of target positions in ra and dec . and convert them to accurate plate x y coordinates using the optical distortion model of the prime focus corrector and the differential atmospheric refraction at the expected time of observation .
these x y positions were then converted to mechanical drawings ( with temperature compensation to allow for the difference between the average day time drilling temperature and the average night time observing temperature ) , and these were the input to a cnc machine for drilling the plates to the required specification .
a schematic of a galaxy field plate is shown in fig .
[ fig : plate ] .
aaomega @xcite is a fibre spectrograph designed for use with the 2-degree field ( 2df ) robotic fibre positioner on the aat .
it has a double - beam schmidt design that allows for optimised performance in the red and blue simultaneously .
volume - phase holographic ( vph ) gratings are employed to give high throughput .
there are a range of gratings available , giving resolutions from @xmath41700 to @xmath413000 for the sami fibres ( 105@xmath47 m diameter ) .
there is a choice of two dichroics that split the light and direct it to the blue and red cameras ; one dichroic has a cut - off wavelength of 570 nm and a second at 670 nm .
the detectors are 2k@xmath464k e2v devices with the short axis in the wavelength direction and the long axis in the spatial direction . for operation with the 2df top - end fibre feed
, the aaomega slit is populated with 392 fibres that each have 140@xmath47 m cores , projecting to @xmath43.4pixels on the detector , with a core - to - core pitch of 10pixels .
aaomega is also fed by the spiral ifu used at the aat cassegrain focus .
spiral has a 16@xmath4632 rectangular lenslet array , with a sampling of 0.7arcsec .
the 512 spiral fibres have 85@xmath47 m core diameters , which project to @xmath42.4pixels on the detector .
the fibre cores are separated on the slit by a pitch of 140@xmath47 m . for sami we have matched the fibre core - to - pitch ratio of spiral , which is @xmath85 .
this is more closely packed than used for 2df , but has proven sufficient with spiral for the minimisation of crosstalk between fibres using an optimal extraction methodology described by @xcite .
the sami fibre pitch on the slit is 170@xmath47 m , giving a core - to - pitch ratio of @xmath86 .
this pitch then defines the maximum number of hexabundles ( 13 ) that will fit along the concave aaomega slit while allowing 26 individual sky fibres to also be positioned on the slit . for convenience of manufacture and assembly ,
each hexabundle and sky fibre pair is fed to an individual silica v - groove slitlet .
the mapping is such that the central core is in the middle of the slit .
the outward spiral numbering of the hexabundle cores corresponds to odd and even outwardly alternating fibres in the slitlet , with the sky fibres being the outermost fibres on the slitlet .
this arrangement maximises the number of adjacent hexabundle cores that are also adjacent on the detector and minimises cross - contamination between sky and adjacent object spectra .
the aaomega slit is mounted on a rotating mechanism that allows four different slits to be positioned in the beam of the spectrograph .
this is used during operation of 2df , when one of the two fields is being observed while the other is being re - configured by the robot .
the sami slit is located at one of the previously spare positions on the aaomega slit rotator ; in the course of normal sami operations the slit does not move .
sami uses an adaption of the 2df / aaomega control software @xcite .
this software has significant flexibility and had been successfully evolved through various changes since it was originally commissioned with 2df in 1996 . as part of the implementation of aaomega in 2006 ,
the software was modified to support the spiral ifu feed to aaomega .
this operational mode did not use the fibre positioner and adc of the 2df top - end , and as a result was a good starting point for sami .
a number of minor modifications were required to the software to support this additional mode of operation .
the 2df / aaomega software inserts a large binary table into the ccd data files to fully describe the location and status of each of the fibres .
this table differs significantly from versions implemented for spiral and 2df .
a component of the software ( the `` fibre to fits '' program ) is used to build this table , and a new version of this program was required for sami .
it reads a file describing the allocation of objects to ifus and sky fibres .
the software combines this with the telescope pointing information and information from files containing the measured positions of each fibre in each ifu to determine the actual location of each fibre on the sky . as a result ,
a euro-3d compliant @xcite binary table can be generated and world coordinate system ( wcs ) information given for each fibre .
extra information in the table is provided to ensure every fibre is fully traceable , and that its position and the origins of its position are well defined .
it is important that fields be observed sufficiently close to the time planned , and for which the plug - plate was drilled .
otherwise changes in airmass can cause some objects to be poorly acquired , largely due to differential atmospheric refraction across the fov . to assist in the decision whether to observe a given plate , the control software displays the error between the drilled hole and the current object sky position for each of the 13 ifu probes at the beginning of each exposure .
data reduction for sami is performed using the 2dfdr data reduction pipeline @xcite originally written for the 2df instrument and then modified for use with the aaomega spectrograph .
this provides fully automated reduction of flat fields , arcs and object frames , including spectral extraction , wavelength calibration and sky subtraction .
the code is run from a fully configurable gui which allows user control of the algorithms used .
the main modifications required for sami were a revised algorithm to accurately map the fibre locations across the detector ( the ` tramline map ' ) and routines to read new elements of the fits binary table in the data frames .
2dfdr is sufficiently fast that data can be reduced in real time , a feature that is of particular value during commissioning of an instrument .
the 2dfdr pipeline generates extracted , flat - fielded , wavelength - calibrated , throughput - normalized and sky - subtracted spectra .
the data product is a 2d image containing each of the fully reduced 1d spectra , together with a variance array and binary table ( see section [ sec : software ] ) . to allow fast reconstruction of 3d data - cubes and first - pass science analysis ,
we have developed a suite of python - based routines which allow us to : ( i ) construct and write full data cubes for individual ifus ( ii ) view all 13 reconstructed ifu images collapsed over a user - specified wavelength range ; ( iii ) calculate the centroid of a source within the ifu field ; ( iv ) calculate the offset in arcseconds between the source and the centre of the ifu ; ( v ) extract summed spectra for an entire ifu or view individual spaxel spectra ; and ( vi ) fit emission lines and construct kinematic maps on the fly . as an aid to the development of the data reduction pipeline ,
an instrument data simulator was used .
this was based on the simulator used for the new high resolution hermes spectrograph being developed for the aat @xcite , but modified to simulate sami detector images .
this was of particular value in developing the algorithm required to extract the fibre spectra from the 2d image .
[ cols="^,<,<,<,>",options="header " , ] the current system , as described above , performs well ( see results below ) , but substantial improvements can be made in performance and usability . to this end
a new fibre cable will replace the current one in the first half of 2012 .
this will substantially improve the system throughput , particularly at wavelengths @xmath87 .
the new fibres will be housed in a low stress cable to further improve throughput by substantially reducing frd .
a third planned modification will be to use a different form of connector for the hexabundles . while providing accurate alignment in the @xmath74 and @xmath75
, the current sma connectors do not allow easy rotational alignment .
the sma connectors will be replaced with a keyed connector such as the fc type .
the astrometric calibration plates consist of four stacked stellar fields ( @xmath88mag ) , each having 13 ifus for distortion mapping and alignment refinement as well as four stacked bright stellar fields ( @xmath89mag ) each with at least two stars for initial visual alignment .
the calibration plate field uses a centred guide star located at the field centre .
the calibration plates are derived from the stellar data contained within the tycho - i reference catalogue .
we trimmed the catalogue on position ( dec.@xmath90 degrees ) and magnitude ( @xmath89 for visual alignment fields , @xmath88 for calibration fields ) for a number of observable ra regions .
the targets for the commissioning were selected to test the full capability of sami , subject to the following criteria : + ( i ) the density of targets on the sky exceeded the 13 hexabundle units available in the 1degree field of sami .
+ ( ii ) the targets were of sufficient angular size to fill the 15 arcsec aperture of each fibre bundle .
+ ( iii ) galaxy surface brightnesses ( in either emission lines or continuum ) were likely to yield s / n@xmath913 in the outermost fibres of each bundle in a @xmath42-hour integration .
we decided to select commissioning targets from a wide cross - section of galaxy spectral types to better constrain the performance of the instrument across a range of future scientific applications .
the target input catalogue was drawn from the 6-degree field galaxy survey ( 6dfgs ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , which itself was selected from the two - micron all - sky survey extended source catalog ( 2mass xsc ; * ? ? ?
the 6dfgs catalogue contains 125071 redshifted southern galaxies ( @xmath92 ) selected to @xmath93 , @xmath94 , @xmath95 , @xmath96 , and @xmath97 .
the median redshift of the sample is @xmath98 .
apparent galaxy sizes are taken from the 2mass xsc .
the mean density of 6dfgs galaxies is @xmath47@xmath99 , substantially lower than the @xmath419@xmath99 ifu density of sami .
however , the low - redshift nature of the sample means that 6dfgs target densities vary considerably over the sky , and in the densest regions , 6dfgs target densities exceed the ifu density of sami . by careful pre - selection of dense regions we ensured that all 13 hexabundles were filled for every field .
nine dense regions ( each of diameter 8deg ) were selected over a range of hour angle and declination to provide 4086 potential targets for the final allocation of sami fields and ifus .
targets were ranked on a scale from 1 ( lowest priority ) to 9 ( highest priority ) that was used to weight the targets for hexabundle allocation and field placement .
targets were given a greater relative weighting if they had : ( i ) large angular size ; ( ii ) prominent spectral emission lines ; or ( iii ) a spectrum typical of an early - type galaxy , with high signal - to - noise .
the angular sizes used in criterion ( i ) were 2mass @xmath100-band half - light radii .
galaxies were divided into those with apparent radial sizes @xmath101 ( @xmath102 ) , those with @xmath103 ( @xmath104 ) , those with @xmath105 ( @xmath106 ) , and the smallest ones that remained , with @xmath107 ( @xmath108 ) .
these divisions were chosen to differentiate between those galaxies most closely matched to the hexabundle aperture size ( with galaxy diameters 1020@xmath109 ) , and those outside this range .
for the purpose of commissioning , the largest galaxies were also given high priority to extend the range of test subjects .
criterion ( ii ) was based on matches to the emission - line galaxy spectral templates used during the 6dfgs cross - correlation redshifting procedure ( see * ? ? ?
note that it is not a complete sample , as spiral galaxies which are dominated by their bulge on the scale of the 6dfgs fibre aperture ( 6.7@xmath109 diameter ) may not exhibit substantial emission lines in their 6df spectrum . criterion ( iii ) was populated by the 6dfgs velocity catalogue ( 6dfgs - v ; * ? ? ?
* ; * ? ? ?
* ) , a high signal - to - noise ( @xmath110 per pixel in the 6dfgs spectrum ) subsample of 11288 early - type galaxies . selecting the target sample in this way ensured that the performance of sami would be gauged across a broad range of galaxy types , while also maximising size and surface brightness considerations , where possible .
table [ priorities ] summarises the selection criteria applied to each of the priority assignments , and the total number of available targets in each case .
the first commissioning observations were carried out on the nights of 14 july 2011 at the aat .
the primary aims were to test the alignment and astrometry of the plug - plate and hexabundle units , estimate system throughput , and examine the reliability and robustness of the instrument .
the secondary goal was to obtain the first galaxy ifu observations with sami to test data quality and determine the accuracy with which physical parameters could be extracted .
the aaomega spectrograph setup used the 580v grating in the blue arm at a central wavelength of 480 nm , covering the wavelength range from 370 nm to 570 nm ( the latter set by dichroic that splits the light between the red and blue arms of aaomega ) . with the 105@xmath47 m fibre cores of sami , the 580v grating
provides a spectral resolution of @xmath111 ( or 173 fwhm ) .
the dispersion is 0.103nmpixel@xmath15 .
this wavelength range is ideal for measuring a wide range of spectral features in the blue parts of galaxy spectra at low redshifts ( @xmath112 ) , such as the d4000 break , emission , various hydrogen balmer lines , and mgb . in the red arm ,
the key spectral features are the , and emission lines .
as these are all located in a relatively narrow wavelength range ( and the target redshift range is small ) , it is possible to observe all of these features using the higher resolution 1000r grating .
this was set to have a central wavelength of 680 nm , providing a spectral range from 625 nm to 735 nm .
the spectral resolution in the red arm is @xmath113 ( or 67 fwhm ) , with a dispersion of 0.057nmpixel@xmath15 .
the first observations used astrometric calibration plates to measure the accuracy of ifu placement , including checks of rotation , plate - scale and distortion .
an initial correction for rotation was applied by one of us ( sr ) observing a small number of bright ( @xmath89 ) stars by eye from within the aat prime focus unit , using calibration holes drilled in the plate at the position of the stars .
the stars were aligned to the holes using the rotational adjustment micrometer on the plate holder assembly ( section [ sec : prime ] ) . once this first correction was made , several astrometric calibration fields
were observed in order to make precise measurements of the required parameters .
after the astrometric observations where carried out , we targeted a number of standard stars to measure system throughput , and then a galaxy field was observed as part of science verification . during periods of poor weather other tests
were carried out , including examination of guide camera flexure and variations in fibre throughput as a function of telescope position .
the detailed results of all commissioning observations are presented in section [ sec : results ] below .
all data presented below was reduced using 2dfdr ( see section [ sec : drsoftware ] ) .
fibre flat field frames illuminated by a quartz lamp are reduced first to define the location of the spectra ( tram - line map ) , and to construct a fibre - flat field , which all other frames are normalized by .
this provides correction of the relative colour response of the fibres , but not total throughput normalization .
extraction of the spectra was carried out using gaussian fits to the fibre profiles , based on a mean profile shape .
wavelength calibration was via an arc frames using a copper - argon lamp .
throughput calibration of the fibres was carried out either by using twilight sky frames or the strength of night sky emission lines in the object spectra .
once the fibres were flat - fielded , wavelength - calibrated and corrected for relative throughput , sky subtraction was performed .
a median sky spectrum was constructed from the 26 sky fibres .
this was then subtracted from the object spectra .
the relative throughput of each fibre is shown in fig [ fig : fibre_thput ] .
these are normalized to the median throughput , so are distributed about a value of 1 .
the full range of throughputs is from @xmath114 to 1.2 .
a marked periodic structure is seen every 8 fibres ( see inset in fig [ fig : fibre_thput ] ) which is due to poorer throughput at the edges of each fibre ribbon ( see section [ sec : frd ] ) and the relative difference is consistent with the expected throughput loss from frd measured in the lab based experiments .
the measured rms variation in relative fibre throughput is 0.13 , however the uniform throughput away from the ribbon edges suggests that a rebuilt fibre cable not using ribbons would provide much more uniform throughput .
good throughput calibration and sky subtraction are crucial in obtaining accurate spectroscopy at faint surface brightnesses . here
we make a preliminary assessment of sky subtraction precision , noting that various upgrades to the data reduction code are still to be implemented .
here we focus on the residual continuum in fibres which were located on blank sky .
minimizing the residual continuum is critical in enabling accurate measurements of stellar population parameters ( e.g. age and metallicity ) from absorption line indices .
the subtraction of night sky emission lines , while also important , can be tackled in a variety of ways , including principle component analysis @xcite that can effectively remove emission to the poisson limit of the data .
this approach will be implemented for sami data .
as the galaxy commissioning targets completely filled the field of view of each hexabundle ( see figs .
[ fig : gals ] and [ fig : kin ] ) , we used only the sky fibres for this test .
it should be noted that examination of the spectral point - spread function ( psf ) of sky and hexabundle fibres found no noticeable difference between them , suggesting that a test of only the sky fibres provides a fair assessment of sky subtraction accuracy .
future on - sky tests will confirm this with blank sky observations using the hexabundles .
we median filter the sky spectra before and after sky subtraction and then take the ratio of the sum of these median filtered spectra as a measure of the fractional residual sky continuum present . the rms continuum residual about zero
is 3.5 percent , suggesting that in the current data this is the level at which we are accurately subtracting sky continuum .
there are a small number ( typically @xmath115 ) sky fibres which perform significantly worse that this ( residuals of @xmath116 percent or more ) .
examination of these fibres showed that i ) they do not demonstrate sky emission line residuals at the same level , and ii ) they tend to be located next to the most badly damaged hexabundle which projects essentially no light onto the spectrograph ccd .
this points to residual scattered light in the spectrograph as the main cause .
in particular , the fibre profiles are made up of a gaussian core and low level , but broad , scattering wings . in regions of the ccd which are fully populated with fibres the scattering wings coadd to form an approximately constant pedestal above the bias level of the detector ( when data are sky limited ) .
however , in detector regions without illuminated fibres the background level falls below the pedestal caused by the scattering wings . as a result
, fibres on the edge of a blank region have a modified psf and can be poorly extracted from the data frame .
the solution to this is to fit a more complex psf to the data in the extraction process ( e.g. * ? ? ?
this will be implemented within the 2dfdr package .
the aat prime focus triplet corrector distortion is dominated by a pincushion effect . for sami
, we implemented a model of this using the sla library routine sla_pcd @xcite and the recommended value of 178.585 for the pincushion coefficient , @xmath117 .
the radial distance in the presence of distortion is given by @xmath118 where @xmath36 is the radial distance from the tangent point .
distances are in units of the projection radius .
an analysis of a zeemax model of the optical system agreed with the previously used model values , to a maximum error of 0.25 arcseconds at the field edge .
careful attention was paid to the many orientation and sign issues throughout the software and plate manufacturing process , leading to the first plate being successfully acquired with only a minor rotation correction required .
-band ( 57507000 ) indicated by the blue horizontal line .
the comparison between directly observed and lab predicted throughputs is approximate as the on - sky observations did not cover the whole of the photometric @xmath71-band.,width=321 ] the aao implemented a package for calibrating the astrometric models in multi - object spectrographs for the eso vlt flames instrument @xcite and later for the subaru fmos instrument @xcite .
this package , known as fpcal , can be configured for new instruments by the use of plug - in software components .
these components provide the program with an implementation of the optical model and the ability to read details of objects allocated to fibres and the telescope astrometric model parameters from the fits files generated by the instrument .
fpcal then provides tools to allow fitting of model parameters and analysis of the results .
the user of fpcal can select either singular value decomposition ( svd ) or powell s method of minimization ( e.g. * ? ? ?
* ) as the fitting algorithm . in practice
, these give the same results within noise margins , so the svd technique is normally chosen as it is faster and provides more information .
the appropriate plug - ins to enable us to calibrate sami with fpcal were implemented .
fpcal requires as an additional input the offset necessary to centre each ifu , and software was implemented in python to extract this from the reduced data files .
centroids were calculated using a gaussian fit to determine the position of calibration stars with respect to the ifu centre .
observations of three different astrometric calibration fields were made , including fields drilled in different physical plates .
an analysis that includes acquisition error , scale , rotation and distortion was performed for each individual field .
sets that combine multiple observations where the stars were offset by @xmath42 fibre cores with respect to the ifu centres were used to test the rotation of individual ifus . in the initial analysis , results were entirely dominated by minor acquisition and scale issues ; the following refers to one typical observation .
an acquisition error of approximately 1.25@xmath109 was removed to simplify further analysis .
a fit to the scale found a focal length about 20 mm longer than originally presumed ; this scale error caused positional errors of up to 1.2@xmath109 at the edge of the plate .
removing the scale error resulted in 0.8@xmath109 rms residuals . fitting the pin - cushion distortion parameter and field rotation did not provide a significantly better fit ( 0.78@xmath109 rms residuals ) . whilst the change in focal length of 20 mm is larger
then expected , this is believed to be due to the power of the triplet corrector , and the actual physical change involved is likely small .
lcccccccccc hexabundle & r.a . & dec .
& priority & 6dfgs i d & redshift & @xmath119 & @xmath120 & @xmath121 & radius@xmath122 & throughput + & ( j2000 ) & ( j2000 ) & & & & & & & ( arcsec ) & flag@xmath40 + h#016 & 195518.8 & -545935 & 1 & g1955188 - 545935 & 0.019 & 15.43 & 1.29 & 4.68 & 11.6 & good + h#015 & 195526.8 & -550920 & 3 & g1955268 - 550920 & 0.045 & 16.49 & 1.43 & 4.30 & 3.4 & good + h#005 & 195651.4 & -545837 & 1 & g1956514 - 545837 & 0.061 & 16.65 & 1.37 & 4.35 & 5.2 & bad + h#010 & 195651.6 & -551956 & 1 & g1956516 - 551956 & 0.056 & 16.35 & 1.23 & 4.23 & 5.3 &
good + h#004 & 195656.7 & -554730 & 1 & g1956567 - 554730 & 0.018 & 14.67 & 1.03 & 3.62 & 8.1 &
good + h#014 & 195711.2 & -552509 & 1 & g1957112 - 552509 & 0.059 & 16.40 & 1.33 & 4.41 & 4.4 & good + h#009 & 195722.2 & -550814 & 1 & g1957222 - 550814 & 0.016 & 15.36 & 1.25 & 4.31 & 10.6 & good + h#011 & 195733.7 & -553441 & 1 & g1957337 - 553441 & 0.017 & 14.45 & 1.26 & 4.32 & 8.1 & good + h#001 & 195800.3 & -553329 & 1 & g1958003 - 553329 & 0.056 & 16.90 & 1.46 & 4.55 & 3.4 & good + h#012 & 195812.8 & -554053 & 9 & g1958128 - 554053 & 0.017 & 15.62 & 0.16 & 4.07 & 14.5 & good + h#006 & 195829.1 & -550913 & 1 & g1958291 - 550913 & 0.058 & 16.14 & 1.40 & 4.65 & 4.9 & bad + h#013 & 195845.0 & -553511 & 1 & g1958450 - 553511 & 0.058 & 16.11 & 1.24 & 4.20 & 5.5 & good + h#002 & 195617.6 & -550711 & 1 & g1956176 - 550711 & 0.060 & 16.96 & 1.73 & 3.97 & 2.4 & bad + + + during the sami commissioning run two spectrophotometric standards were observed over two nights in several of the individual ifus . on 2 july 2011
the star ltt6248 was observed in ifu hexabundles 10 and 13 ( h#010 and h#013 ) .
the first of these observations was used to calculate the overall throughput of the sami system including the atmosphere , telescope primary , prime focus corrector , sami fibre feed , and aaomega spectrograph .
first an integrated spectrum was extracted from the ifu data by summing the spectra in all spatial elements of the ifu .
since the data are under - sampled and the ifu grid is not contiguous , some light is lost between spaxels and a summed spectrum can under - estimate the true flux in the star by as much as 10 percent .
this can be avoided by performing a more careful psf extraction of the star over the relevant fov , but in this case we used the simpler method and this should be borne in mind for the following results .
the summed spectrum was then compared to the tabulated values for the spectrophotometric standard .
the results of this analysis are shown in figures [ fig : tp_b ] and [ fig : tp_r ] for the blue and red arms of the spectrograph respectively .
these throughput curves have not been corrected for telluric features , which are still visible .
the curves therefore show the measured throughput of the entire system from atmosphere to detector , including the aat and the aaomega spectrograph . of note
is the fact that the throughput falls off quite quickly in the blue arm .
this is not ideal but the source of the effect is known .
the fibre train for sami consists of @xmath442 m of afs105/125y fibre matched to the fibre used in the manufacture of the hexabundles themselves .
this type of fibre has poorer throughput in the blue than the polymicro fbp fibre used , for example , in 2df @xcite .
throughput values measured from the standard star were compared with the lab - tested throughput from section [ sec : frd ] .
this is a difficult comparison because the lab tests use bessel @xmath70-band and @xmath71-band filters and the throughput measured from the standard star observations changes rapidly through the bandpass of the bessel @xmath70-band filter .
the lab tests were corrected for the known throughput in the @xmath70- and @xmath71-bands of the atmosphere ( 0.72 , 0.89 for @xmath70 and @xmath71 respectively ) , telescope including corrector ( 0.77 , 0.80 ) , and spectrograph including ccd ( 0.17 , 0.33 ) .
the lab tests have uncertainties due to fibre - to - fibre differences .
the data reduction includes fibre - to - fibre throughput corrections for the standard stars , which scales the throughput of each individual fibre to be closer to that of the fibre in the centre of a ribbon .
we therefore compare the throughput measured from the observations to core 6 in table [ frd ] .
the sami standard star images had a steep variation in observed throughput across the @xmath70-band ranging from 2646 percent ( for sami only ) , while the lab - tests found @xmath65 percent .
the @xmath71-band lab measurement is not directly comparable to the on - sky measurement , as the @xmath71-band filter used in the lab has a peak transmission wavelength outside the range of the aaomega 1000r grating .
however , we show an indicative comparison between direct and lab based throughput measurements in fig . [
fig : tp_r ] . the throughput comparisons , while limited by the differences in beam shape and filters used , shows that there are no other significant losses in the fully installed system that are not accounted for in the above analysis of throughput . from the measured throughput of the sami system we can calculate the limiting surface brightness for typical observations , assuming photon counting errors .
here we assume 3 hours exposure time per field and a dark sky ( @xmath84 and @xmath71 band sky brightnesses of 21.5 and 20.8 magsarcsec@xmath123 ) . for a s
/ n of 5 @xmath15 the estimated limiting surface brightness is 22.9 and 22.5 magarcsec@xmath123 for the @xmath84 and @xmath71 bands respectively ( on the vega scale ) . for a fiducial galaxy survey sample , which is flux limited to @xmath124 and redshift @xmath125
, these surface brightnesses limits will allow the current sami system to achieve s / n@xmath126 at 1 @xmath127 for 80 percent of galaxies . 6548 , 6583 and @xmath1286716 , 6731 emission lines are clearly visible.,width=321 ]
5007 are clearly visible in the integrated and central spectra but have lower s / n in the outer spectrum .
the masked regions in the spectra are due to bad pixels in the aaomega blue ccd.,width=321 ] one galaxy field with thirteen galaxies was observed during the commissioning run .
three 40-minute exposures were taken at this pointing on the night of 2 july .
these were reduced using 2dfdr and stacked , resulting in a total of 2 hours integration on this field .
a spectrophotometric standard star was observed on the same night as the objects and was reduced in an identical way .
the extracted star spectrum was then used to correct the galaxy data for instrument throughput , though the data were not fully flux - calibrated .
the details of the 13 target galaxies are given in table [ tab : targets ] .
figure [ fig : gals ] shows the ten galaxies from the observed field that had usable data . the remaining 3 hexabundles were damaged during commissioning .
the images were created by summing over a wide wavelength range ( 63007300 in the observed frame ) ; the colour bars indicate the logarithm of the flux in arbitrary units .
the discs and bulges of the larger galaxies are readily apparent representative spectra of the galaxy observed in bundle h#009 are shown in figures [ fig : spec_red ] and [ fig : spec_blue ] .
the three spectra in each figure correspond to the integrated spectrum of the galaxy ( top panel ) , the spectrum observed in the central core ( core 1 ; the central panel ) and the spectrum observed in an outer core ( core 28 ; the bottom panel ) .
the integrated spectra are simply the sum of all the individual spaxel spectra in a particular ifu .
various emission and absorption features are seen in the galaxy spectra . in the red wavelength range the , @xmath1286548 , 6583 and @xmath1286716 , 6731 emission lines are clearly visible in all three spectra , though with lower s / n in core 28 .
in the blue wavelength range we see the and @xmath75007 emission lines , with an absorption feature also visible .
the data in core 28 have lower s / n than the central core , but still sufficient for analysis . , width=321 ] in examining the results , we first studied the integrated spectrum of each galaxy .
seven galaxies show medium to strong and emission ( along with the doublet at 6716 and 6731 ) and three of these also show the and lines . the ratios of these lines are often used to classify objects by means of a so - called bpt diagram @xcite .
each line was fitted with a gaussian to measure the line strength . in the case of
, a weak absorption trough was observed in the summed spectra of all three galaxies ; this trough was fitted separately and accounted for when calculating the line strength .
figure [ fig : bpt_all ] shows the galaxies on a bpt diagram ( large diamonds ) . in figure
[ fig : bpt_all ] the blue curve shows the theoretical maximum line for star - forming galaxies after @xcite while the red curve shows the empirical cut - off for star - forming galaxies from @xcite .
the black line is an empirical line , also from @xcite , that separates agn ( above the line ) from liners ( below the line ) .
the galaxy h#009 ( an image of which is shown in the left - hand panel of figure [ fig : h9_all ] ) was then studied in more detail , using the full spatially - resolved set of spectra .
the line ratios discussed above were calculated for each individual spaxel in the ifu fov , yielding spatially - resolved line ratio maps , which are shown in the centre and right panels of figure [ fig : h9_all ] .
there is a systematic spatial trend in both line ratios moving away from the plane of the galaxy s disc .
the green circles in figure [ fig : bpt_all ] shows the line ratios from each spaxel plotted on a bpt diagram .
a strong trend is seen in the sense that the spaxels further from the plane of the disc have a harder ionising mechanism , reflecting the spatial trend seen in the line ratio maps .
a single - fibre spectrum of this galaxy would have seen only the central few arcseconds of the object and resulted in classification as a normal star - forming galaxy .
it is clear from the sami results , however , that there is a more interesting story behind this galaxy , which will be discussed in detail in a future paper ( fogarty et al . in preparation ) .
gas kinematics were examined for six of the galaxies using the h@xmath129 emission line . in each spaxel of each galaxy the h@xmath129 line was fitted by a gaussian function and the kinematic properties ( mean velocity and velocity dispersion ) extracted from the fit .
the resulting velocity maps are shown in figure [ fig : kin ] .
the colour bars show velocity in . from these maps
it is clear the galaxies h#009 , h#010 , h#012 and h#016 are fairly typical rotating discs ; the velocity maps for h#001 and h#004 are possibly more complex . from the wealth of spatially - resolved information presented for the handful of galaxies studied in this single field during the commissioning run ,
it is clear that sami can explore a very wide parameter space with enormous science potential .
in this paper we have presented a new instrument , the sydney - aao multi - object ifs ( sami ) .
sami makes use of astrophotonic technology in the form of hexabundles ( multi - core fibre bundles ) to enable simultaneous ifu observations of 13 objects over a 1degree diameter accessible field .
each ifu contains 61 elements , each 1.6 arcsec in diameter , giving a fov for each ifu of 15 arcsec .
sami has now been commissioned on the anglo - australian telescope ( aat ) and we demonstrate its science potential via preliminary observations of galaxies selected from the 6df galaxy survey .
we make the case that multiplexed integral field spectroscopy is the natural next step in galaxy surveys , which to date have been dominated by multiplexed single - aperture observations .
the extra information gained by ifu spectroscopy , combined with the statistical power of a large survey , will enable a fundamental step forward in our understanding of galaxy formation and evolution by distinguishing the spectroscopic properties of the major structural components of galaxies .
assuming that a survey could observe 3 fields per night ( i.e. nominal exposure times between 2 and 3hours ) , sami will allow 10,000 galaxies to be targeted in 260 clear nights on the aat .
much greater gains could be made with an increased fov and larger numbers of ifus .
a key requirement for making full use of the larger fields of view available on current and future telescopes is to have sufficient spectrographs to handle all the fibres from the ifus .
a 2degree diameter fov ( e.g. that available with the aat s 2df corrector ) could accommodate @xmath450 or more ifu bundles comparable to those in sami ; this would , however , require @xmath43300 fibres ( including sky fibres ) .
such large numbers of fibres naturally drive designs towards mass - produced fixed - format spectrographs , such as those currently being developed for projects such as muse @xcite and virus @xcite
. more development work on the bundle technology is under way .
one possible improvement is to force the fibre cores into a perfect hexagonal configuration to simplify the data analysis , particularly when using dithering to smooth out the discrete sampling .
this will not be easy to do with our current core size while at the same time ensuring an ar - coated polished front facet that is not overly stressed by the hexagonal grid . with the successful commissioning on the aat of fibre - based oh suppression technology @xcite
, we can envisage whole bundles that are fully sky - suppressed at near - infrared wavelengths .
once the near - infrared sky is rendered as dark as the optical sky , exposure times in both regimes are expected to be the same and it makes sense to run optical and infrared spectrographs simultaneously .
we thank the staff of the australian astronomical observatory and those at the university of sydney for their excellent support in developing , construction and commissioning of the sami instrument .
smc acknowledges the hospitality of the leibniz institute for astrophysics in potsdam ( aip ) during the completion of this paper .
we thank jakob walcher , martin roth , roger haynes and davor krajnovi for helpful discussions .
the centre for all - sky astrophysics is an australian research council centre of excellence , funded by grant ce11e0090 .
smc acknowledges the support of an australian research council ( arc ) qeii fellowship ( dp0666615 ) , an australian research council future fellowship ( ft100100457 ) and an j g russell award from the australian academy of science .
jbh is supported by a federation fellowship from the arc .
cqt gratefully acknowledges support by the national science foundation graduate research fellowship under grant no .
dge-1035963 .
r. et al .
, 2004 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 5492 , society of photo - optical instrumentation engineers ( spie ) conference series , a. f. m. moorwood & m. iye , ed .
, pp . 11451149 m. , smedley s. , barnes s. , farrell t. , barden s. , 2010 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
7735 , society of photo - optical instrumentation engineers ( spie ) conference series g. j. , macqueen p. j. , tejada c. , cobos f. j. , palunas p. , gebhardt k. , drory n. , 2004 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
5492 , society of photo - optical instrumentation engineers ( spie ) conference series , a. f. m. moorwood & m. iye , ed .
, pp . 251261 g. a. et al .
, 2004 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 5492 , society of photo - optical instrumentation engineers ( spie ) conference series , a. f. m. moorwood & m. iye , ed .
, pp . 410420 d. q. , cui x. , wang y. , yao z. , 1998 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 3352 , society of photo - optical instrumentation engineers ( spie ) conference series , l. m. stepp , ed . , | we demonstrate a novel technology that combines the power of the multi - object spectrograph with the spatial multiplex advantage of an integral field spectrograph ( ifs ) .
the sydney - aao multi - object ifs ( sami ) is a prototype wide - field system at the anglo - australian telescope ( aat ) that allows 13 imaging fibre bundles ( `` hexabundles '' ) to be deployed over a 1degree diameter field of view .
each hexabundle comprises 61 lightly fused multimode fibres with reduced cladding and yields a 75 percent filling factor .
each fibre core diameter subtends 1.6 arcseconds on the sky and each hexabundle has a field of view of 15 arcseconds diameter .
the fibres are fed to the flexible aaomega double beam spectrograph , which can be used at a range of spectral resolutions ( @xmath0 170013000 ) over the optical spectrum ( 37009500 ) .
we present the first spectroscopic results obtained with sami for a sample of galaxies at @xmath1 .
we discuss the prospects of implementing hexabundles at a much higher multiplex over wider fields of view in order to carry out spatially resolved spectroscopic surveys of @xmath2 galaxies .
instrumentation : spectrographs techniques : imaging spectroscopy surveys galaxies : general galaxies : kinematics and dynamics |
convolutional networks ( convnets ) have recently enjoyed a great success in large - scale image and video recognition @xcite which has become possible due to the large public image repositories , such as imagenet @xcite , and high - performance computing systems , such as gpus or large - scale distributed clusters @xcite . in particular , an important role in the advance of deep visual recognition architectures has been played by the imagenet large - scale visual recognition challenge ( ilsvrc ) @xcite , which has served as a testbed for a few generations of large - scale image classification systems , from high - dimensional shallow feature encodings @xcite ( the winner of ilsvrc-2011 ) to deep convnets @xcite ( the winner of ilsvrc-2012 ) . with convnets becoming more of a commodity in the computer vision field , a number of attempts have been made to improve the original architecture of @xcite in a bid to achieve better accuracy .
for instance , the best - performing submissions to the ilsvrc-2013 @xcite utilised smaller receptive window size and smaller stride of the first convolutional layer .
another line of improvements dealt with training and testing the networks densely over the whole image and over multiple scales @xcite . in this paper
, we address another important aspect of convnet architecture design its depth . to this end
, we fix other parameters of the architecture , and steadily increase the depth of the network by adding more convolutional layers , which is feasible due to the use of very small ( @xmath1 ) convolution filters in all layers . as a result ,
we come up with significantly more accurate convnet architectures , which not only achieve the state - of - the - art accuracy on ilsvrc classification and localisation tasks , but are also applicable to other image recognition datasets , where they achieve excellent performance even when used as a part of a relatively simple pipelines ( e.g.deep features classified by a linear svm without fine - tuning ) .
we have released our two best - performing models to facilitate further research .
the rest of the paper is organised as follows . in sect .
[ sec : arch_config ] , we describe our convnet configurations .
the details of the image classification training and evaluation are then presented in sect .
[ sec : learning ] , and the configurations are compared on the ilsvrc classification task in sect .
[ sec : exp ] .
[ sec : conclusion ] concludes the paper . for completeness
, we also describe and assess our ilsvrc-2014 object localisation system in appendix [ sec : loc ] , and discuss the generalisation of very deep features to other datasets in appendix [ sec : dataset_transfer ] .
finally , appendix [ sec : revisions ] contains the list of major paper revisions .
to measure the improvement brought by the increased convnet depth in a fair setting , all our layer configurations are designed using the same principles , inspired by @xcite . in this section ,
we first describe a generic layout of our configurations ( sect .
[ sec : arch ] ) and then detail the specific configurations used in the evaluation ( sect . [
sec : config ] ) .
our design choices are then discussed and compared to the prior art in sect .
[ sec : discuss ] . during training , the input to our convnets is a fixed - size @xmath2 rgb image .
the only pre - processing we do is subtracting the mean rgb value , computed on the training set , from each pixel .
the image is passed through a stack of convolutional ( conv . )
layers , where we use filters with a very small receptive field : @xmath1 ( which is the smallest size to capture the notion of left / right , up / down , center ) . in one of the configurations we also utilise @xmath3 convolution filters , which can be seen as a linear transformation of the input channels ( followed by non - linearity ) .
the convolution stride is fixed to @xmath4 pixel ; the spatial padding of conv .
layer input is such that the spatial resolution is preserved after convolution , i.e.the padding is @xmath4 pixel for @xmath5 conv . layers .
spatial pooling is carried out by five max - pooling layers , which follow some of the conv .
layers ( not all the conv
. layers are followed by max - pooling ) .
max - pooling is performed over a @xmath6 pixel window , with stride @xmath7 . a stack of convolutional layers ( which has a different depth in different architectures )
is followed by three fully - connected ( fc ) layers : the first two have 4096 channels each , the third performs 1000-way ilsvrc classification and thus contains 1000 channels ( one for each class ) .
the final layer is the soft - max layer .
the configuration of the fully connected layers is the same in all networks .
all hidden layers are equipped with the rectification ( relu @xcite ) non - linearity .
we note that none of our networks ( except for one ) contain local response normalisation ( lrn ) normalisation @xcite : as will be shown in sect .
[ sec : exp ] , such normalisation does not improve the performance on the ilsvrc dataset , but leads to increased memory consumption and computation time .
where applicable , the parameters for the lrn layer are those of @xcite . the convnet configurations , evaluated in this paper , are outlined in table [ tab : config ] , one per column . in the following
we will refer to the nets by their names ( a e ) .
all configurations follow the generic design presented in sect .
[ sec : arch ] , and differ only in the depth : from 11 weight layers in the network a ( 8 conv . and 3 fc layers ) to 19 weight layers in the network e ( 16 conv . and
3 fc layers ) . the width of conv .
layers ( the number of channels ) is rather small , starting from @xmath8 in the first layer and then increasing by a factor of @xmath7 after each max - pooling layer , until it reaches @xmath9 . in table
[ tab : num_params ] we report the number of parameters for each configuration . in spite of a large depth , the number of weights in our nets is not greater than the number of weights in a more shallow net with larger conv .
layer widths and receptive fields ( 144 m weights in @xcite ) . a & a - lrn & b & c & d & e + 11 weight & 11 weight & 13 weight & 16 weight & 16 weight & 19 weight + layers & layers & layers & layers & layers & layers + + conv3 - 64 & conv3 - 64 & conv3 - 64 & conv3 - 64 & conv3 - 64 & conv3 - 64 + & * lrn * & * conv3 - 64 * & conv3 - 64 & conv3 - 64 & conv3 - 64 + + conv3 - 128 & conv3 - 128 & conv3 - 128 & conv3 - 128 & conv3 - 128 & conv3 - 128 + & & * conv3 - 128 * & conv3 - 128 & conv3 - 128 & conv3 - 128 + + conv3 - 256 & conv3 - 256 & conv3 - 256 & conv3 - 256 & conv3 - 256 & conv3 - 256 + conv3 - 256 & conv3 - 256 & conv3 - 256 & conv3 - 256 & conv3 - 256 & conv3 - 256 + & & & * conv1 - 256 * & * conv3 - 256 * & conv3 - 256 + & & & & & * conv3 - 256 * + + conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 + conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 + & & & * conv1 - 512 * & * conv3 - 512
* & conv3 - 512 + & & & & & * conv3 - 512 * + + conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 + conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 & conv3 - 512 + & & & * conv1 - 512 * & * conv3 - 512 * & conv3 - 512 + & & & & & * conv3 - 512 * + + + + + + [ tab : config ] .*number of parameters * ( in millions ) . [ cols="<,^,^,^,^,^,^",options="header " , ] [ tab : generalise_action ] our representation achieves the state of art on the voc action classification task even without using the provided bounding boxes , and the results are further improved when using both images and bounding boxes . unlike other approaches
, we did not incorporate any task - specific heuristics , but relied on the representation power of very deep convolutional features .
[ [ other - recognition - tasks . ] ] other recognition tasks .
+ + + + + + + + + + + + + + + + + + + + + + + + since the public release of our models , they have been actively used by the research community for a wide range of image recognition tasks , consistently outperforming more shallow representations .
for instance , @xcite achieve the state of the object detection results by replacing the convnet of @xcite with our 16-layer model .
similar gains over a more shallow architecture of @xcite have been observed in semantic segmentation @xcite , image caption generation @xcite , texture and material recognition @xcite .
here we present the list of major paper revisions , outlining the substantial changes for the convenience of the reader . | in this work we investigate the effect of the convolutional network depth on its accuracy in the large - scale image recognition setting . our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small ( @xmath0 ) convolution filters , which shows that a significant improvement on the prior - art configurations can be achieved by pushing the depth to 1619 weight layers .
these findings were the basis of our 2014 submission , where our team secured the first and the second places in the localisation and classification tracks respectively .
we also show that our representations generalise well to other datasets , where they achieve state - of - the - art results .
we have made our two best - performing convnet models publicly available to facilitate further research on the use of deep visual representations in computer vision . |
the field of molecular electronics was arisen by the early realization that organic molecules could act as rectifiers@xcite when attached to conducting electrodes to form tunnel junctions .
many experiments with a large variety of organic molecules have been performed@xcite , typically finding values of the conductance @xmath1 several orders of magnitude smaller than @xmath0 ( @xmath2 is the conductance quantum ) and a large variability , which hinder the reproducibility of the experiments .
molecular junctions can be understood in terms of resonant tunneling models@xcite , where the conduction is carried through the highest occupied and lowest unoccupied molecular orbitals ( homo and lumo , respectively ) .
these are revealed as sharp resonances in either the densities of states ( dos ) of the molecule , or the transmission coefficients @xmath3 of the junction , and are usually located 1 or 2 ev above or below the fermi level of the molecule , respectively .
conductance values of the order of @xmath0 can only be achieved by pinning one of those resonances to the fermi level of the electrodes .
otherwise , the conductance is very low .
the invention of the scanning tunneling microscope@xcite allowed the fabrication of stable atomic point contacts@xcite .
these junctions were found to be highly transparent in many cases , and to show values of the conductance of the order of @xmath0 , which confirmed early theoretical predictions on the matter@xcite .
theoretical analyses of these junctions found that their transmission coefficients @xmath3 show wide plateaus as a function of the energy @xmath4 of the incoming electrons , with heights of order one .
the high transparency of these junctions is due to the good matching between the conduction channels at the electrodes and those at the molecule .
the ( contact ) resistance of the junction is different from zero because of the different number of channels at the electrodes and the junction which leads to a recombination of the former to match the later@xcite .
importantly , more recent developments using mcbj techniques demonstrated that high values of the conductance ( @xmath5 ) are not restricted to atomic constrictions , but could also be obtained even when platinum or palladium electrodes are bridged by hydrogen molecules@xcite .
these results were reproduced by theoretical simulations@xcite , which also showed how the antibonding level of the hydrogen molecule hybridized strongly with the conduction channels of the electrodes , providing a junction with a single conduction channel .
the transparency of this channel was manifested in the transmission coefficients @xmath3 , that had a wide plateau of height one .
furthermore , very recent experimental work confirms that junctions comprising platinum electrodes and either simple benzene@xcite , or a number of small molecules@xcite show conductance values of the order of @xmath0 . as stated above
, junctions in the tunneling regime can shed conductance values of order @xmath0 provided that a molecular level is exactly pinned to the fermi energy .
but it is hard to believe that all of the junctions discussed in those recent experiments@xcite display this pinning mechanism .
instead , they clearly indicate that highly transparent molecular junctions can be fabricated with relative ease .
they also indirectly hint that thiol - capping necessarily leads to junctions in the tunneling regime .
we have performed a number of transport simulations of molecular junctions where first - row diatomic molecules are sandwiched by semi - infinite ( 001 ) gold and platinum electrodes , with our code smeagol@xcite .
our simulations confirm that this type of junctions is highly conductive . indeed
, the transmission coefficients do not show resonant behavior but wide plateaus of height of order one instead . to sustain theoretically these simulations
, we argue that the conductive behavior of a junction is determined by two factors .
first , by the conjugation nature of the molecule , e.g. : whether the homo and lumo levels are delocalized throughout the whole molecule , or not .
second , by the strength of the chemical bond between the conduction channels at the electrodes and these delocalized homo or lumo orbitals
. a junction will be highly conductive provided that its molecule is conjugated and the chemical bond referred above is strong .
conjugated molecules are reasonably well approximated by caroli s model@xcite .
we use this model below to show that when the chemical bond between the delocalized homo or lumo orbitals and the conduction channels is strong , the transmission coefficients show wide plateaus whose height is of the order of @xmath0 .
we argue that these plateaus are robust against changes in the energy of the homo / lumo levels , so that there is no need to fine - tune their position to the fermi level of the electrodes in order to achieve large conductance values . on the contrary ,
when the conduction channels at the electrodes do not bind chemically with neither the homo nor the lumo levels then the junction is in the tunneling regime where it shows resonant behavior .
diatomic or triatomic molecules are sufficiently small to show conjugation . likewise
, benzene - based molecules are archetypical conjugated molecules , where the conjugation is driven by internal @xmath6-bonding .
it is therefore important to understand unequivocally why bdt junctions show strong tunneling behavior while on the contrary the simpler benzene junctions are highly transparent .
we perform below a study of the transmission coefficients of these junctions .
our study indicates that direct au - s bonding is detrimental of their conductive behavior .
we have performed simulations of the conductance of ( 001 ) gold and platinum junctions that are bridged by diatomic molecules of the first - row elements , using our code smeagol@xcite .
smeagol is a transport program which uses the non equilibrium green s functions formalism@xcite to compute the charge density and current of the junction .
the hamiltonian of the junction is determined by dft theory via the code siesta@xcite . in the case of gold junctions
, we have taken for simplicity a flat surface , oriented the molecules perpendicular to it , and set the electrode - molecule distance equal to the equilibrium distance of the gold - h@xmath7 junction .
we have used a more realistic geometry for platinum in contrast , motivated by the recent experiments by tal and coworkers@xcite .
we have placed in these cases a pyramid of platinum atoms on top of the flat surfaces , as in our earlier publications@xcite .
we have oriented the molecules in a variety of angles , including the bridge , perpendicular and tilted orientations .
we have finally relaxed the forces of the atoms at the pyramids and molecules .
we have found in this respect that li@xmath7 , be@xmath7 , b@xmath7 , c@xmath7 and n@xmath7 relax to the bridge position , while o@xmath7 , f@xmath7 and co relax to a tilted orientation .
our code computes the conductance of the junction via the formula @xmath8 where @xmath9 are the energy- and voltage - dependent transmission coefficients of the junction , defined in eq .
( 17 ) in appendix a , and @xmath10 are the distribution functions of the left and right electrodes , also defined in appendix a. @xmath11 can be approximated at low enough voltages by the linear response formula @xmath12 we show in figs . 1 and 2 the transmission coefficients @xmath3 for
all the gold and platinum junctions simulated .
notice that we have taken the fermi energy of the semi - infinite electrodes at equilibrium as the reference energy .
overall , we find that the zero - voltage conductance @xmath13 is of order @xmath0 , with the only exception of the gold-@xmath14 junction .
further , all the transmission coefficients are rather smooth at positive energies , while they show peaks in a range of energies below @xmath15 .
these peaks do not correspond to molecular states , but rather to the d - band conduction channels at the electrodes , and are naturally located either a few @xmath16 below @xmath15 for gold , or extend up to @xmath15 for platinum .
we will skip below any further reference to this d - band conduction channels , since they are not relevant for our purposes
. a closer look at fig .
1 shows that @xmath17 for gold junctions is either a plateau ( as in li@xmath7 , be@xmath7 and f@xmath7 ) , or the sum of a plateau and a broad resonance ( in b@xmath7 , c@xmath7 , n@xmath7 o@xmath7 and c ) .
the low - voltage conductances are close to @xmath0 for li@xmath7 and f@xmath7 , while they are of order 0.3 - 0.7 @xmath0 for be@xmath7 , c@xmath7 , n@xmath7 and co , and of order 1.5 - 2 @xmath0 for b@xmath7 and o@xmath7 .
the transmission coefficients of platinum junctions in fig .
2 show a slightly more complex structure , but are still pretty smooth and show no sign of the narrow resonances that mark the tunneling regime . for platinum , li@xmath7 , and n@xmath7 show conductances of about @xmath0 , while be@xmath7 and co have about half a conductance quantum , and b@xmath7 , c@xmath7 and o@xmath7 have conductances in the range 1.5 - 2 @xmath0 .
the conductance of f@xmath7 is below 0.1 @xmath0 .
our results for co agree with a previous simulation preformed by strange and coworkers@xcite , but fail to reproduce the peak at 1 @xmath0 shown in the experimental conductance histograms@xcite . as a side remark , we should point out that we have found that the conductance of platinum may increase or decrease by even a factor of two depending on the placement and orientation of the molecule .
caroli s model in its simplest form , is depicted schematically in fig .
it consists of two identical semi - infinite chains , called left ( l ) and right ( r ) electrodes , which sandwich a free - standing atom ( m ) , so that the three of them are initially unconnected .
the chains have a single orbital per atom of energy @xmath18 ; electrons can hop between atoms via the hamiltonian matrix element @xmath19 .
the three pieces are initially held at the same bias voltage , and the common fermi energy @xmath15 of the unbiased chains is taken as the reference energy . since the chains are semi - infinite , they can be regarded as reservoirs of electrons , subjected to a given equilibrium chemical potential @xmath20 , whose energy level population is described by the fermi distribution functions @xmath10 .
the central atom is held initially in thermodynamic equilibrium by its contact to a third reservoir at the arbitrary chemical potential @xmath21 .
this determines the equilibrium distribution of the atom @xmath22 and therefore its population .
the whole system is subsequently biased by a voltage @xmath23 so that the chemical potentials of the chains is shifted to @xmath24 , while the chemical potential at the atom can still be left equal to the initial value , although it is physically more reasonable to reset it to the average between @xmath25 and @xmath26 .
notice that the three pieces stay initially in equilibrium since they are unconnected .
later on , the central atom is connected to the electrodes by adiabatically switching on the hopping integrals between them until they reach their final value @xmath27 .
we assume that the system is able to reach an stationary non - equilibrium state long time afterwards@xcite , where there is a total bias voltage between the electrodes equal to @xmath23 , that induces a finite electron current .
it is important to stress that the charge population at the atom is determined by its contact to three reservoirs , each at a different chemical potential .
notice furthermore that , while the coupling to the chains is controlled by @xmath27 , the coupling to the third reservoir can not be modulated , so that the flow of electrons between it and the atom is completely transparent .
this model can easily be solved analytically using the machinery of the non - equilibrium green s functions formalism@xcite ( see appendix a for a thorough algebraic derivation ) .
4 shows the results for the density of states at the atom ( dos ) @xmath28\ ] ] where @xmath29 is the retarded green s function at the central atom , which is defined in appendix a. the figure also shows the conductance @xmath11 , which has been computed in this case performing the numerical derivative of the current through the junction , to access the high - voltage behavior .
notice that in this model the position of the homo / lumo levels with respect to the fermi energy of the electrodes ( taken as 0 ) is given by the atomic energy @xmath30 , while the hopping integral @xmath27 denotes the strength of the homo / lumo hybridization with the electrodes .
the hybridization causes two important and well - known effects on the bare atomic level @xmath30 .
first , it renormalizes it , e.
g. : the level changes its energy .
second , its broadens it , since now an electron initially placed at the atomic level can hop back and forth to the electrodes , and hence the atomic state acquires a finite lifetime .
these two effects are readily seen in @xmath31 , whereby the initial delta - like peak corresponding to the atomic states moves and broadens to a resonance .
the resonance is sharp if @xmath27 is much smaller than @xmath19 , or broad if @xmath27 and @xmath19 are of the same order of magnitude .
what we wish to stress here is that the extent of this broadening also determines the behavior of the conductance of the junction .
4 illustrates how the different regimes of a junction manifest in the dos and the conductance and how these regimes are controlled by the basic parameters of the model .
the top three panels correspond to the resonant tunneling regime .
notice that this regime is seen even for values of the hybridization @xmath27 as large as 0.2 .
this regime is characterized by a sharp resonance in the dos , located at the energy position of the molecular orbital @xmath30 .
additionally , sharp resonances appear in the conductance at different voltages , whose maximum height can not exceed @xmath0 .
the low - voltage conductance is very small except if the molecular orbital is pinned to the fermi energy , in which case @xmath32 .
we stress that the molecular level must be fine - tuned to the fermi energy to achieve sizable values of the conductance .
it is important to stress again that the resonant tunneling regime is achieved by values of @xmath27 which are smaller , _ but not much smaller _ , than @xmath19 .
just a factor of five is enough to place a junction within it .
the bottom panels in fig . 4 correspond to highly transparent junctions , which corresponds to values of @xmath33 .
the molecular orbital is in this case fully hybridized with the conduction channels at the electrodes .
the dos is very broad , and has a width of the order of the bandwidth of the conduction channels at the electrodes .
more importantly , the conductance is very flat and its height is very close to @xmath0 . notice that this happens for a wide range of positions of the molecular orbital @xmath30 .
therefore , if the hybridization between the molecular orbital and the conduction channels is large , then there is no need whatsoever to fine - tune the position of the molecular orbital . in other words : when @xmath33 , the molecular orbital is always tuned , provided it is initially located within the band of the conduction channels at the electrodes .
this is one of the central results in this article .
the middle panels in fig . 4
show the situation for junctions of intermediate transparency , the dos reflects this crossover behavior , where a resonance has already been developed but still has a large width .
the conductance is very flat , but we point that its height has a stronger dependence on the position of the molecular level than was the case of the highly transparent junction .
it is fortunate that some _ ab initio _
simulation codes@xcite allow to make approximate estimates of the physical parameters that appear in caroli s model .
we have indeed recently performed such estimates for a constriction consisting of platinum electrodes bridged by @xmath34 molecules .
we found that in this case @xmath35 ( e.g. : @xmath36 ) , and that the anti - bonding state of the molecule was located within the s - d - band complex of platinum@xcite .
we hence ascribed that junction to the highly transparent regime , driven by the large value of @xmath27 .
most of the junctions that we have simulated in this article , on the other hand , correspond to the crossover regime , where the conductance is of order @xmath0 , but where its exact value has a significant dependence on the exact position of the molecular orbital .
this is also the case of the benzene junctions discussed by kiguchi and coworkers@xcite . to make a closer contact with experiments on diatomic molecules ,
we have performed a slight modification of caroli s model .
that we call the diatomic model . in it ,
the central atom is replaced by a diatomic molecule as depicted in fig . 1 ( b ) .
the nice feature of this model is that on the one hand , it accounts for both homo and lumo levels and , on the other , it is also easily solved analytically ( the algebra is relegated to appendix b ) .
this model can be applied directly to the case of the platinum-@xmath34 constriction , where we showed that the electronic conduction is carried by the anti - bonding state of the molecule , which is strongly hybridized to the platinum conduction bands , while the bonding state lies slightly below the edge of the platinum conduction band and hence does not participate in the chemical bond , showing up as a sharp resonance in the dos@xcite .
as stated above , we estimated that for this case @xmath37 ( @xmath36 ) , while @xmath38 was slightly larger , about 6 ev ( @xmath39 ) .
we display in fig .
5 the results of the diatomic model for the dos at the left atom in the molecule , and for @xmath11 for the case where @xmath40 , and where the bonding state lies below the conduction band edge and hence shows up as a sharp resonance in the dos .
we again find that when @xmath27 is much smaller than @xmath19 , both the dos and @xmath1 display resonant behavior , whereby the conductance is always very small , except for specific ( usually too large ) voltages , or when the anti - bonding state is finely positioned at the fermi energy . on the contrary ,
when @xmath41 , the resonance broadens , and the conductance has long plateaus with values close to @xmath0 .
the results that compare best to our simulation data for the dos at the molecule in the platinum-@xmath34 constriction ( see fig . 3 in ref . [ ] ) correspond to the middle panel in the bottom row .
hence we expect that this model can describe correctly the platinum-@xmath34 constriction if we use the parameters @xmath42 , @xmath43 and @xmath44 .
notice that the conductance indeed displays a value close to @xmath0 .
a final note in this section relates to the large variability in the conductance obtained experimentally in bdt junctions .
we remind that the voltage at which the conductance resonances depends strongly on the energy position of the molecular orbital for junctions belonging to the resonant tunneling regime . in other words ,
the value of the measured conductance depends strongly on the details of the orbital and its bonding for this type of junctions . on the contrary , since the conductance has a weak dependence on the details of the molecular orbital for highly transparent junctions , we expect that the experimental variability must be suppressed for them .
benzene and benzene - dithiolate are conjugated molecules whose homo and lumo states correspond to @xmath6-bonded p orbitals , which are delocalized throughout the whole molecule .
bdt / gold structures are archetypical molecular junctions , which are used as templates against which both theory and experiments are benchmarked .
experimental data show always small conductance values , which range from one to three orders of magnitude smaller than @xmath0 , depending on the experiment .
theoretical results tend to shed larger values for @xmath1 , typically of the order of 0.05 to 0.1 @xmath0 .
this discrepancy is usually attributed to a poor description of electronic correlations by the theoretical models@xcite .
bdt / gold junctions therefore seem to belong to the resonant tunneling regime described in the previous section .
interestingly , benzene / platinum junctions are highly conductive@xcite , showing values of @xmath1 of the order of @xmath0 , which indicates that these junctions belong to the transparent regime of caroli s model .
it is therefore important to understand better the contrasting behavior of these seemingly similar junctions . in order to do so
, we have simulated bdt junctions attached to ( 001 ) flat gold surfaces .
bdt molecules attach via the thiol - groups , such that the sulfur atoms are placed in hollow sites at each side , as shown in fig .
we have set a sulfur / gold distance of 2 , such that the junction is slightly stretched .
we have also simulated benzene molecules attached to ( 001 ) gold surfaces .
the surfaces are initially flat , but we have placed an additional gold atom on top of each of them , to provide for a preferred anchoring site .
benzene molecules bind to these electrodes via the @xmath6-conjugated molecular orbitals , yielding a geometry for the junction such as that shown in fig .
we have set a distance of 5.0 between the two apex gold atoms , which is again slightly longer than the equilibrium distance of the junction .
we note that there is no direct hybridization between gold orbitals at each side of the junction for such a distance , which is confirmed by negligible transmission coefficients @xmath3 . upon relaxation of forces
, we have found that the center of the benzene plane is tilted in such a way to maximize the bonding between the @xmath6 orbitals and the apex gold atoms .
the transmission coefficients of both junctions are plotted in figs . 7 ( a ) and ( b ) .
we find that they both show broad resonances , which classifies them in a crossover ground between the resonant tunneling and intermediate transparency regimes .
consequently , the effective hybridization between the homo / lumo levels and the conduction channels is neither too large nor too small .
we therefore expect that the conductance of these junctions must have a relatively strong dependence on the details of the junctions .
bdt junctions have two main peaks peaks , placed 1 ev below and 2 ev above the fermi level , respectively .
there is a gap in between , so that our simulations yield a zero - voltage conductance of about 0.1 @xmath0 .
this height is actually reduced by strong correlation effects@xcite .
the placement of the two peaks and the gap is due to the fact that the @xmath6-conjugated orbitals are orthogonal to the conduction channels of the electrodes , due to the geometry of the junction .
the geometry of benzene / gold junctions on the contrary is such that the @xmath6-conjugated orbitals hybridize directly with the conduction channels at the electrodes .
consequently , the broad resonance is moved , or pinned , to the fermi level .
the reactivity of the thiol groups is therefore detrimental of the conductivity of bdt junctions . in other words ,
bdt molecules attach to the electrodes via the thiol groups , yielding a junction geometry where the conjugated orbitals do not bind to the conduction channels .
when the thiol groups are missing , benzene attaches to the electrodes via the less reactive @xmath6 molecular orbitals .
while the @xmath6-gold chemical bond is less reactive , it has the virtue of providing a direct hybridization between the homo / lumo orbitals and the conduction channels at the electrodes .
we have performed simulations of a number of molecular junctions , where ( 001 ) gold or platinum electrodes sandwich first - row diatomic molecules .
we have found that these junctions are highly conductive , which is manifested both in large values of the conductance and in smooth transmission coefficients @xmath3 .
we have used caroli s model to argue that this is a generic feature of the transparent regime of a junction , which is driven by a high hybridization between the delocalized molecular orbitals in the molecule and the conduction levels at the electrodes .
jf acknowledges conversations with j. van ruitenbeek , as well as his sharing his results with us prior to submission .
this research has been funded by the spanish government ( project fis2006 - 12117 ) .
it is convenient from a mathematical point of view to split the total hamiltonian in two pieces : the unperturbed hamiltonian of the total system @xmath45 and the perturbation that drives the system out of equilibrium @xmath46 .
we use the basis of atoms in the system , whose @xmath47 states we denote by @xmath48 .
then the hamiltonians can be written as the following matrices @xmath49 where @xmath50 are @xmath51 tridiagonal matrices of the form @xmath52 and @xmath53 and @xmath54 are @xmath55-dimensional vectors .
the retarded and advanced green s functions of the unperturbed system can be calculated through the equations : @xmath56\,{\cal g}^{r , a}_0(w)={\cal i}\ ] ] where @xmath57 , @xmath58 being an infinitesimal number .
notice that @xmath59 are also huge matrices of size @xmath60 .
this large set of coupled equations can actually be reduced enormously by gaussian elimination of the atoms in the electrodes until only three states remain , @xmath61 .
the resulting 3x3 matrix green s functions are called @xmath62 to avoid confusing them with the conductance , which is denoted by @xmath1 .
their matrix elements ( r / a superindices are henceforth dropped when there is no danger of confussion ) @xmath63 here @xmath64 are the surface green s function of the electrodes , and in our one - dimensional model they are equal to @xmath65 to simplify matters , we will take henceforth @xmath66 and @xmath19 as the energy unit .
we also define a _ lesser _ green s function , that carries information of the electron occupation in each piece , @xmath67 where each @xmath68 is written in terms of the density of states @xmath69 and the fermi distribution function @xmath70 of each unconnected piece , as @xmath71 .
then the keldish formalism@xcite provides a simple recipe to compute the green s functions of the final system in the steady state , @xmath72^{-1}-{\cal h}_1]^{-1}= \left(\begin{array}{ccc } g_{ll}^r & g_{lm}^r & g_{lr}^r \\
g_{ml}^r & g_{mm}^r & g_{mr}^r
\\ g_{rl}^r & g_{rm}^r & g_{rr}^r \end{array}\right ) \\\\ & f^<=g^r\,[g^r_0]^{-1}\,g^<_0\,[g^a_0]^{-1}\,g^a= \left(\begin{array}{ccc } g_{ll}^ < & g_{lm}^ < & g_{lr}^ < \\
g_{ml}^ < & g_{mm}^ < & g_{mr}^ < \\
g_{rl}^ < & g_{rm}^ < & g_{rr}^ < \end{array}\right ) \end{array}\ ] ] these equations can easily be solved analytically , with the following result for the retarded / advanced green s function , @xmath73 where @xmath74 and @xmath75 . likewise , @xmath76 can be written as the sum of the following three matrices @xmath77 @xmath78 @xmath79 these lesser green s functions enter the calculation of the electronic charge and current .
for instance , the charge at the atom @xmath80 can be written as follows : @xmath81 this formula for the charge can be rewritten in terms of the conventional expression for the equilibrium state @xmath82\,n_l\ ] ] plus an explicit non - equilibrium term @xmath83 , and likewise @xmath84 .
the current traversing the left contact is computed via the formula @xmath85 which can be written as a sum of the following two contributions , @xmath86 notice that the current traversing the right contact can be written as the sum of @xmath87 plus @xmath88 where @xmath89 since these atomic contributions apparently break the conservation of charge , @xmath90 are usually dropped ( as is the second term in eq . ( 15 ) ) .
in other words , the total current @xmath91 is approximated by @xmath87 .
the unperturbed hamiltonian and the perturbation in the diatomic model take the form : @xmath92 where @xmath50 are the @xmath51 tridiagonal matrices described in appendix a , and @xmath93 after decimating the unwanted degrees of freedom , the unpertubed green s functions look like @xmath94 where @xmath95 and @xmath96 .
the algebra is more tedious , but the final result for the retarded green s function @xmath97 is the following : @xmath98 & t_m\,\sigma_l\ , ( \omega-\varepsilon_m-\sigma_r ) & t_m\,t\,\sigma_l & t\,\sigma_l\,\sigma_r \\
t_m\,\sigma_l\,(\omega-\varepsilon_m-\sigma_r ) & t_m^2\,(\omega-\varepsilon_m-\sigma_r ) & t_m^2\,t&t_m\,t\,\sigma_r \\ t_m\,t\,\sigma_l & t_m^2\,t & t_m^2\,(\omega-\varepsilon_m-\sigma_l ) & t_m\,\sigma_r\,(\omega-\varepsilon_m-\sigma_l ) \\
t\,\sigma_l\,\sigma_r & t_m\,t\,\sigma_r & t_m\,\sigma_r\,(\omega-\varepsilon_m-\sigma_l ) & \sigma_r\,[(\omega-\varepsilon_m)(\omega-\varepsilon_m-\sigma_l)-t^2 ] \end{array}\right)\ ] ] where @xmath99 , and the equation for the current traversing the left link can be written as @xmath100 where the atomic contribution has been neglected again , and @xmath101 refers to the element ( 2,3 ) of the retarded green s function matrix , that connects the two atoms in the molecule . | we present a theoretical study of the transport characteristics of molecular junctions , where first - row diatomic molecules are attached to ( 001 ) gold and platinum electrodes .
we find that the conductance of all of these junctions is of the order of the conductance quantum unit @xmath0 , spelling out that they belong to the transparent regime .
we further find that the transmission coefficients show wide plateaus as a function of the energy , instead of the usual sharp resonances that signal the molecular levels in the tunneling regime .
we use caroli s model to show that this is a rather generic property of the transparent regime of a junction , which is driven by a strong effective coupling between the delocalized molecular levels and the conduction channels at the electrodes .
we analyse the transmission coefficients and chemical bonding of gold / benzene and gold / benzene - dithiolate ( bdt ) junctions to understand why the later show large resistances , while the former are highly conductive . |
we have recently developed a robust and accurate algorithm to track vortex lines in the gross - pitaevskii equation ( gp ) with arbitrary geometries in a periodic domain .
the full details of the algorithm and the case studies to check its validity can be found in @xcite .
we recall here briefly the basic ideas .
a quantised vortex line is defined by the nodal lines of the wavefunction . in three dimensions
this corresponds to a line defined by @xmath67=im[\psi(x , y , z)]=0\ ] ] the algorithm is based on a newton - raphson method to find zeros of @xmath2 and on the knowledge of the _ pseudo - vorticity _ field @xmath68\times \nabla im[\psi]$ ] , always tangent to the line , to follow vortex lines ( in the spirit of rorai et al .
@xcite ) . starting from a point @xmath69 where the density @xmath70 is below a given small threshold ( therefore very close to a vortex ) , we define the orthogonal plane to the vortex line using @xmath71 .
the plane is then spanned by the two directors @xmath72 and @xmath73 as illustrated in fig.[fig : sketch ] . ] a better approximation for the true point @xmath74 where the vortex lies on the plane is then given by @xmath75 . here the increment @xmath76 is obtained using the newton - raphson formula : @xmath77 where @xmath78 is the jacobian matrix expressed as @xmath79\cdot \hat{{\bf u}}_1 & \nabla re[\psi]\cdot \hat{{\bf u}}_2 \\ \nabla im[\psi]\cdot \hat{{\bf u}}_1 & \nabla im[\psi]\cdot \hat{{\bf u}}_2 \end{pmatrix}.\ ] ] the increment can be therefore found using @xmath80,im[\psi({\bf x_0})])^t$ ] . the jacobian matrix is a non - singular @xmath81 matrix so its inverse is computed with its explicit formula .
we underline that the method will require in general the evaluation of the jacobian at intermesh points .
however , making use of the spectral representation of @xmath2 , we can compute spatial derivatives of the field @xmath82 efficiently using fast fourier transforms .
this process can be iterated until the true vortex location @xmath74 is determined upon a selected convergence precision . to track the next vortex point we use as next initial guess @xmath83 , which is obtained evolving along @xmath84 by a small step @xmath85 .
the process is repeated until the entire line is tracked and closed .
once the line is completely tracked the process is repeated with another line until the whole domain has been fully explored .
the algorithm has been extensively described and validated in @xcite using different case tests . as a final remark
, we underline that the algorithm takes full advantage of the spectral resolution when a field is generated integrating the gross - pitaevskii with a pseudo - spectral method .
a comparison between the low - value isosurfaces of the density field and tracked vortices is displayed in fig.[fig : tangle ] for different times of the taylor - green decay .
.,title="fig:",scaledwidth=45.0% ] .,title="fig:",scaledwidth=45.0% ] .,title="fig:",scaledwidth=45.0% ]
in this work we use the so - called taylor - green initial condition introduced by nore et al . @xcite .
we recall here for the sake of completeness how this initial condition is produced making use of the clebsch representation of a vector field .
let us assume that an incompressible velocity field @xmath86 admits a global clebsch representation in terms of the potentials @xmath87 and @xmath88 @xcite , such that @xmath89 .
it follows that @xmath90 the clebsch representation has an interesting geometrical representation .
a vortex line of the velocity flow @xmath91 is mapped into a point in the @xmath92 plane .
indeed a vortex line @xmath93 is defined by @xmath94 .
a simple solution is given by @xmath95 and @xmath96 , with @xmath97 and @xmath98 constants .
it follows that , using a @xmath99 wavefunction @xmath100 of a vortex at the origin , it is straightforward to construct a @xmath101 wavefunction with a nodal line defined by @xmath102 and @xmath103 .
this @xmath101 field simply reads @xmath104 this idea is applied to the taylor - green velocity flow defined in the domain @xmath105 ^ 3 $ ] as : @xmath106 it is composed of 8 fundamental boxes that are obtained one from each other by mirror symmetric transformation of the sub - box @xmath107 ^ 3 $ ] .
it admits a decomposition in terms of the clebsch potentials ( see @xcite for further details ) : @xmath108 in the sense that @xmath109 .
note that the circulation around the plane @xmath110\times[0,\pi]$ ] is @xmath111 a set of vortex lines can be defined in the wavefunction as @xmath112 any @xmath99 vortex profile approximation can be used as it will be later relaxed by using the advective - real - ginzburg - landau equation @xcite . in order to match ( as close as possible ) the circulation to the taylor - green flow
, the initial condition is defined as @xmath113 where @xmath114 and @xmath115 represent the quantum of circulation .
the wavefunction @xmath116 is then evolved under the advective - real - ginzburg - landau equation : @xmath117 subject to the chemical potential @xmath118 .
this equation is just the imaginary time evolution of the gross - pitaevskii equation after a galilean transformation with a non - constant velocity field @xmath119 .
the final states contains a clean ( without sound ) initial condition with a set of rings ( of charge one ) distributed along the vortical lines of the taylor - green velocity flow .
a visualisation of the taylor - green flow is displayed in fig.[fig : tangle]a .
the initial condition is then evolved using the gross - pitaevskii equation .
the total crossing @xmath120 , total linking @xmath31 , writhe @xmath32 and center - line helicity @xmath40 were computed directly performing the line integrals over the vortex ring(s ) @xcite .
they are defined as @xmath121 @xmath122 @xmath123 where @xmath124 and @xmath125 correspond the points identifying the @xmath126th ring @xmath127 . for the writhe number , @xmath128 and @xmath129
correspond to two different points of the same ring . in fig.[fig :
complexity ] we plot the evolution of the total number of vortex rings and the average crossing number during the development and decay of the tg tangle .
each quantity is normalised with respect to its initial values . as expected from the analysis of the evolution of the total vortex length presented in the letter , both the number of rings and the average crossing number experience an initial increment .
this behaviour can be interpreted as the growth of the complexity in the tangle . ,
@xmath130 and @xmath131.,scaledwidth=45.0% ] as described in the letter , the linking number takes integer values and provides information of the number of links . during the evolution of the taylor - green flow
, rings can link as presented in fig.[fig : rings]a and fig.[fig : rings]b .
these two links were found directly computing the integral for @xmath132 for all pair of rings at each time step .
the writhe number takes real values and provide information of self - linking and helical excitations on the filament . a ring with high value of writhe
is displayed in fig.[fig : rings]c .
note that the writhe number is not enough to determine whether a ring is self - linked ( knotted ) or not .
( 100,400 ) ( -0,280 ) and b ) at time @xmath133 .
c ) a ring with high writhe number at @xmath134,title="fig : " ] ( -0,160 ) and b ) at time @xmath133 .
c ) a ring with high writhe number at @xmath134,title="fig : " ] ( -0,20 ) and b ) at time @xmath133 .
c ) a ring with high writhe number at @xmath134,title="fig : " ] ( -0,380)a ) ( -0,280)b ) ( 0,160)c ) total link and writhe numbers define the center - line helicity which is a measure of the complexity of the vortex tangle considered .
in order to obtain the kelvin wave ( kw ) spectrum of a ring we proceed as follows . starting from a ring @xmath135 provided by the tracking algorithm
we obtain a long - scale averaged ring that will be used to determine the amplitude of the kws . in order to do that ,
we first compute the natural parametrisation @xmath9 of the ring and its total length @xmath136 .
then we apply a gaussian kernel of width @xmath137 , with @xmath138 .
the obtained convolved ring @xmath139 is smooth and it can now be used to define the kws on the ring as @xmath140 by definition , @xmath141 is a periodic set of 3 signals ( one for each spatial dimension ) , hence , the kw spectrum is then defined as @xmath142 where @xmath143 is the fourier transform of @xmath144 . in @xcite
we checked that this procedure is able to capture well the kws superimposed on a ring by performing a series of different tests ( different @xmath145 and amplitude of kws ) . in this work
we use a value of @xmath146 , the variation of this fraction only slightly modifies ( as expected ) the large - scale values of the spectrum , but values in the inertial range remain unchanged .
note that data from the tracking is not spaced in regular intervals of the natural parametrisation . in order to computed the fourier transform , the data is re - meshed in a regular partition of @xmath147 $ ] obtained by a high order interpolation . in reference
@xcite we have also test this technique by comparing it with the method used in @xcite to detect kws in a almost straight vortex line and verified that this interpolation does not affect the obtained spectrum .
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 _ _ , vol .
( , ) link:\doibase 10.1103/physrevb.38.2398 [ * * , ( ) ] link:\doibase 10.1007/s10909 - 011 - 0405 - 6 [ * * , ( ) ] @noop * * , ( ) @noop _ _ , vol .
( , ) link:\doibase 10.1103/physrevlett.103.045301 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/s0167-2789(01)00304-9 [ * * , ( ) ] link:\doibase 10.1103/physrevb.64.134520 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1098/rspa.1957.0191 [ * * , ( ) ] http://dx.doi.org/10.1134/s002136401008014x [ * * , ( ) ] , @noop * * , ( ) @noop ( ) , @noop * * , ( ) link:\doibase 10.1103/physrevlett.100.245301 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.060501 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.075301 [ * * , ( ) ] link:\doibase 10.1073/pnas.1407232111 [ * * , ( ) ] , http://arxiv.org/abs/1602.06880 [ ( ) ] http://dx.doi.org/10.1038/nphys3679 [ * * , ( ) ] link:\doibase 10.1103/physreva.92.063632 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.014502 [ * * , ( ) ] http://stacks.iop.org/1751-8121/47/i=46/a=465101 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.92.035301 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.86.3080 [ * * , ( ) ] link:\doibase 10.1103/physreve.86.055301 [ * * , ( ) ] http://dx.doi.org/10.1038/nphys2734 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.115.170402 [ * * , ( ) ] @noop _ _ ( , ) | the development and decay of a turbulent vortex tangle driven by the gross - pitaevskii equation is studied . using a recently - developed accurate and robust tracking algorithm , all quantised vortices
are extracted from the fields .
the vinen s decay law for the total vortex length with a coefficient that is in quantitative agreement with the values measured in helium ii is observed .
the topology of the tangle is then studied showing that linked rings may appear during the decay .
the tracking also allows for determining the statistics of small - scales quantities of vortex lines , exhibiting large fluctuations of curvature and torsion .
finally , the temporal evolution of the kelvin wave spectrum is obtained providing evidence of the development of a weak - wave turbulence cascade .
the full understanding of turbulence in a fluid is one of the oldest yet still unsolved problems in physics .
a fluid is said to be turbulent when it manifests excitations occurring at several length - scales . due to the large number of degrees of freedom and the nonlinearity of the governing equations of motion ,
the problem is usually tackled statistically by introducing assumptions and closures in terms of correlators .
this is the case in the seminal work of kolmogorov in 1941 based on the idea of richardson s energy cascade , where energy in classical fluids is transferred from large to small scales @xcite .
superfluids form a particular class among fluids characterised essentially by two main ingredients : the lack of dissipation and the evidence that vortex circulation takes only discrete values multiple of the quantum of circulation @xcite .
superfluid examples which are routinely created in laboratories are superfluid liquid helium ( he ii ) and bose - einstein condensates ( becs ) made of dilute alkali gases . here
the superfluid phase is usually modelled via a complex field describing the order parameter of the system and quantised vortices appear as topological defects where the superfluid density vanishes . in three spatial dimensions
those defects organise themselves into closed lines ( or even open lines at the boundaries if confining sides are considered ) of different configurations .
any vortex line point induces a velocity field in the superfluid which affects the motion of any object in the system including the vortex line itself . in general , even for a single closed vortex line , the dynamics are chaotic and the problem does not have analytical solutions .
superfluid turbulence regards the study of the evolution of many vortex lines , a tangle , which induce velocity field gradients in the fluid at several length scales .
different mathematical models have been devised to mimic the dynamics of a superfluid .
an example is the vortex filament ( vf ) model based on the biot - savart law that relates vorticity and velocity @xcite .
this model is able to mimic the dynamics of dense vortex tangles due to a relatively fast numerical integration technique @xcite .
the vf model implicitly assumes that the superfluid density field is constant everywhere and the vortex structure is precisely a line with vanishing core .
this assumption is generally satisfied in he ii where the characteristic experimental setup sizes , and consequently the largest scales of the motion , are order of @xmath0 m and the vortex core is order of @xmath1 m . moreover , since helium is in its liquid phase , the compressibility effects can be usually neglected
. however , the vf model fails to describe vortex reconnections .
these are rapid changes in the topology of the vortex configuration which occur naturally in a superfluid @xcite and are one of the main mechanisms responsible for the energy transfer .
reconnections are thus introduced by some ad - hoc mechanism . another superfluid model that admits quantised vortices and inherently possesses vortex reconnections is the gross - pitaevskii ( gp ) equation that describes the evolution of the superfluid order parameter @xmath2 .
in contrast to the vf model , the gp equation allows density fluctuations in terms of phonons and density depletion at the vortex cores .
although it has been formally derived as a mean - field theory for a dilute boson gas in the limit of zero temperature @xcite , it also qualitatively reproduces he ii dynamics .
the vortex core size in the gp equation is order of the healing length @xmath3 , the only intrinsic characteristic length - scale of the model ; nowadays experimental techniques are able to create bec setups that are @xmath4 healing lengths where superfluid turbulence can develop @xcite . in turbulent superfluids , vortices
constantly re - arrange themselves following reconnections into complex tangles with non - trivial geometrical , algebraic and topological properties @xcite . at small scales , helical excitations of vortex lines known as kelvin waves ( kws ) , are believed to be the ultimate mechanism of energy dissipation via phonon emission @xcite . to study such dynamics , the gp equation has the advantage that no extra modelling is needed ( unlike the vf model ) .
however gp does not provide direct information on vortices .
in this letter we apply a novel numerical algorithm @xcite to track accurately the configuration of a turbulent vortex tangle evolving accordingly to the gp model .
we focus on the evolution and decay of the tangle .
firstly , we show that after the onset of turbulence , the vortex line density satisfies the vinen s decay law @xcite with a coefficient that is in agreement with he ii .
different algebraic and topological quantities of the tangle are then measured .
the tracking allows for obtaining curvature and torsion distributions of the vortex tangle .
finally , we perform a direct measurement of kws during the dynamics and compute a kw spectrum that appears to be consistent with the lvov - nazarenko theoretical prediction @xcite . the gp model for the condensate wave - function @xmath2
is given by @xmath5 where @xmath6 is the mass of the bosons and @xmath7 , with @xmath8 the @xmath9-wave scattering length .
madelung s transformation @xmath10}$ ] relates the wave - function @xmath2 to a superfluid of density @xmath11 and velocity @xmath12 .
the quantum of circulation about the @xmath13 vortex lines is @xmath14 .
when eq . is linearised about a constant value @xmath15 , the sound velocity is given by @xmath16 with dispersive effects taking place at length scales smaller than the healing length @xmath17 .
in the simulations presented in this letter , the mean density is fixed to the unity and the physical constants in eq .
are determined by the values of @xmath18 and @xmath19 .
the quantum of circulation results in @xmath20 .
numerical integration of eq .
is performed using a standard pseudo - spectral code .
we integrate the so - called taylor - green flow @xcite with no enforced symmetries at resolutions @xmath21 and @xmath22 with @xmath23 and @xmath24 respectively ( see appendix b ) . with these units
the largest eddy turnover time is order of the unity .
the taylor - green flow initially contains a configuration of unstable large - scale rings that develop to create a turbulent tangle .
vortices can be easily spotted by plotting the low - value iso - surfaces of the density field as displayed in fig.[fig : snapshot]a .
low - density regions corresponding to vortex lines are plotted in red , while density fluctuations ( sound ) are rendered in light blue .
rendered with vapor software .
low - density regions corresponding to vortex lines are plotted in red , while density fluctuations ( sound ) are rendered in light blue .
b ) corresponding vortex tracking .
different colors correspond to different vortices .
resolution @xmath21.,scaledwidth=45.0% ] we track the vortex lines forming the tangle with a recently - developed algorithm @xcite .
vortex lines are followed using the pseudo - vorticity field @xcite and the exact vortex position is obtained by a newton - raphson method .
the algorithm is robust and accurate as it takes full advantage of the spectral resolution of the field .
it allows for identifying separately each single line forming the tangle .
figure [ fig : snapshot]b shows the corresponding tracked tangle displaying in different colors all the 496 vortex rings . during the decay ,
vortices radiate phonons that populate the small - scales creating a thermal bath that exchanges energy and momentum with the vortices .
this process mimics mutual friction and leads eventually to the total annihilation of vortex rings @xcite . in superfluids such
a decay is modelled by the vinen equation @xcite for vortex line density @xmath25 : @xmath26 where @xmath27 is a constant of order of unity .
its solutions manifest a @xmath28 behaviour at long times : this power - law decay has been named quantum turbulent decay and measured in he ii experiments @xcite and bs numerical simulations @xcite . in fig.[fig : evolutionlinequantities]a we show the temporal evolution of @xmath25 .
( see text ) , together with vinen s prediction @xmath29^{-1}$ ] , with @xmath30 ( solid black line ) .
b ) temporal evolution the total linking @xmath31 , writhe @xmath32 and center - line helicity @xmath33 .
c ) visualisation of two linked rings at @xmath34 .
resolution @xmath21.,scaledwidth=49.0% ]
it is worth noticing that it grows at the initial stages : this is caused by the instability of the initial taylor - green configuration and the subsequent vortex stretching due to numerous vortex reconnections .
the data is compared with an estimation of @xmath35 obtained by computing the ratio between the volume of points such that @xmath36 and the corresponding surface of a perfect two - dimensional vortex profile .
this latter method is a standard technique within gp to estimate @xmath25 numerically @xcite .
even if this technique is able to capture the qualitative behaviour of @xmath35 , it fails to grasp at long times the power - law predicted by vinen s equation .
this is show in the inset of fig.[fig : evolutionlinequantities]a where the measured @xmath37 , setting @xmath38 , is compared to vinen s prediction .
this discrepancy can be explained by the presence of sound waves and small scale kws not detected by the estimation .
the tracking data also allow for determining the numerical constant @xmath39 .
this value is in remarkable agreement with experimental values measured in he ii in the low temperature limit @xcite . from fig.[fig :
tangle ] in appendix a , it is clear that the complexity of tangle first increases and then decreases .
this observation is intuitively explained by the idea that vortex lines simultaneously stretch , bend and coil during reconnection events .
the complexity of tangle can be measured by computing the changes in some of its algebraic and topological quantities , such as the total number of rings , the average crossing number , the total linking @xmath31 and writhe @xmath32 @xcite .
explicit definitions are given in appendix c. the total number of rings and the average crossing number vary , as expected , in a similar manner to the total vortex length in fig.[fig : evolutionlinequantities]a ( see fig.3 in the appendix c ) . here
we focus on the central line helicity @xmath40 @xcite .
this quantity is the analogous to helicity in classical fluid dynamics , an important inviscid invariant .
the linking number @xmath41 takes integer values and gives information about the number of linked rings present in the system , whereas the writhe takes real values and its contribution comes from knots and kws @xcite .
figure [ fig : evolutionlinequantities]b shows the temporal evolution of these three quantities .
initially , @xmath42 , as expected for the taylor - green flow .
surprisingly , during the evolution @xmath41 becomes non - zero , indicating the presence of linked rings , such as the ones displayed in fig.[fig : evolutionlinequantities]c . for the taylor - green flow
] this is remarkable as in the gp model sufficiently simple vortex configurations usually decay by reducing their linking number @xcite .
once the decay is established , no linked rings are present and only writhe contributes to @xmath43 .
the center - line helicity fluctuates about a zero mean , an indication of the presence of kws in this turbulent tangle , as indirectly observed in @xcite .
we now study statistical properties of some geometrical quantities of the vortex filaments .
we explore the time behaviour of the probability density functions ( pdfs ) of the curvature @xmath44 and torsion @xmath45 . in fig.[fig : statistics]a we present the pdf of curvature , normalised by its mean value , at different stages .
normalised by their respective mean values @xmath46 at different times ( same legend as ( b ) ) .
the inset displays the temporal evolution of the mean and rms values of @xmath47 .
b ) pdfs of torsion @xmath48 at different times .
the inset emphasises their @xmath49 power - law tail .
resolution @xmath21.,scaledwidth=49.0% ] the temporal evolution of the mean curvature @xmath50 and its _ rms _ value @xmath51 are also displayed in the inset .
we can observe that @xmath50 increases rapidly at early stages and then almost saturates .
the rms value of the curvature presents the same tendency with the exception of peaks .
these are evidences of reconnection events where high values of curvature are found in localised regions .
it is worth noticing that the pdfs , rescaled by their mean curvature , exhibit a relatively good collapse to a self - similar form .
this latter observation indicates a power - law behaviour @xmath52 at small curvature values , while an exponentially - decaying tail is present at large curvature values .
a similar behaviour has also been observed within the vf model @xcite . in fig.[fig : statistics]c we plot the torsion pdfs at the same stages .
the mean torsion is always about zero and there is no evidence of any skewness in the pdfs .
the distributions tails show an universal power - law behaviour of @xmath53 at all times , meaning that the second and higher moments of the torsion diverge during the decay .
the same scaling appears in vortex tangles of random wave fields that are solutions of the helmholtz equation @xcite .
this might be an indication that for one - time small - scale quantities , quantum turbulent tangles could be interpreted as random vortices .
the large curvature fluctuations and the torsion fluctuation about a zero mean suggest also the presence of kws at all scales propagating on quasi - planar vortex rings . by exploiting the accuracy of the tracking algorithm
we are able to directly detect kws on those rings .
competing theories have been put forward to statistically predict a power - law kw spectrum in the form of @xmath54 ( here @xmath55 is the kelvin wavenumber ) and to explain the energy transfer through kw scales . assuming small amplitude kws ( weak non - linearity ) , kozik&svistunov @xcite and lvov&nazarenko @xcite obtained the exponents @xmath56 and @xmath57 respectively considering two different orders of interaction .
we compute the kw spectra of the @xmath58 largest rings during the evolution of the tangle applying a gaussian kernel in order to establish the unperturbed ring ( see appendix d ) .
the spectra , averaged over the rings , are shown for different times in fig.[fig : kwspectra]a . .
b ) kw spectrum at @xmath59 ( averaged over the 50 longest rings ) for run at resolution @xmath22 .
the dashed line displays the @xmath60 scaling .
the inset displays the respective @xmath61 ( solid blue ) and @xmath62 ( dashed red ) compensated spectra.,scaledwidth=45.0% ] it is evident that all accessible kw modes get populated at early times due to reconnection events that trigger the cascade @xcite .
we observe kw spectra exhibiting power - laws with an exponent independent of time where the best scaling is appreciated at the time where the rings are the longest ( @xmath63 ) . to get the best estimation of the power - law exponent
, we repeated the taylor - green decay in a simulation box twice larger ; in this new configuration the scaling range spans almost two wavenumber decades .
in fig.[fig : kwspectra]b we show the spectrum at @xmath64 : the observed power - law exponent agrees with the lvov&nazarenko @xmath57 prediction
. this can be better appreciated by looking at the compensated spectra with respect to @xmath65 and @xmath66 showed in the inset .
this finding corroborates the result in favour of lvov&nazarenko s prediction previously obtained while studying the kw oscillations about a perfect straight line in the gp model @xcite .
we highlight that although the weak - wave turbulence prediction for the kw spectrum was formally derived for kws on an isolated straight vortex line using the vf model , it remarkably turns out to be valid in a dense turbulent tangle also driven by the gp model .
this is certainly due to the fact that the predicted kw spectrum was found for the longest rings .
small rings quickly loose their energy by phonon radiation and also exchange momentum with sound waves . both contributions are important to understand dissipation of superfluids at very low temperature and further studies are still needed to fully comprehend the relevance of such mechanisms . tracking vortices in gp turbulence opens a new way for studying and understanding the topological configuration and properties of quantum vortex tangles . although unlikely , we show that rings can link creating a local ( in time and space ) fluctuation of helicity .
it will be of great interest to repeat a similar analysis in a gp setting where the mean helicity of the flow is not zero , like the abc flow introduced in @xcite where linking and self - linking processes could be substantially enhanced .
overall , the results presented in this letter confirm that some predictions traditionally associated to superfluid liquid helium become important in weakly - interacting becs at low temperature described by the gp model .
nowadays bec experimentalists are able to create and track few vortices in harmonic traps @xcite . a controlled experimental setting with a turbulent bec , such as the one presented in this letter , has yet to be achieved but it should be realisable in the near future .
gk , dp and av were supported by the cost - share royal society international exchanges scheme ( ref .
ie150527 ) in conjunction with cnrs .
computations were carried out on the msocentre sigamm hosted at the observatoire de la cte dazur and on the high performance computing cluster supported by the research and specialist computing support service at the university of east anglia . |
the fisher information matrix ( fm , @xcite ) is a statistical instrument of paramount importance in parameter estimation problems .
its function is to _ predict _ the precision with which the parameters of interest can be measured by a given experiment , starting from the expected observational uncertainties on the data .
the utility of the fm comes from the fact that it can be used to optimize the experimental setup , according to the desired target results , _ before _ actually collecting the data .
the cramr - rao inequality @xcite states that the uncertainties predicted by the fm formalism are , in general , optimistic ; they set the upper limit for the precision that an experiment can attain .
in particular , the fm method is expected to be accurate only if the probability distribution is a gaussian function of the parameters .
therefore , in order to validate the fm method for a specific problem , it is necessary to compare the predictions obtained via the fm to the results obtained by using other statistical techniques that do not suffer from this limitation . in many fields of astronomy ,
the uncertainties predicted by the fm have been shown to closely resemble the ones estimated by fitting the probability distribution function ( pdf ) .
this is the case , for example , of cosmological parameter estimation from the cosmic microwave background ( @xcite ) , or distance measurements from baryon acoustic oscillations ( @xcite ) ; see also @xcite for a review . in this paper
, we study the applicability of the fm formalism to the problem of fitting the spectral energy distribution ( sed ) of galaxies . in this case
, the planned experiment consists of acquiring multi - wavelength photometry of galaxies , the uncertainties on the projected data are the photometric errors on the flux measurements in each band ( including any systematics ) , and the parameters to be measured are the physical properties of the target galaxies , such as age , stellar mass and dust content . our objective is to demonstrate , for a variety of cases , that the uncertainties forecasted by the fm are comparable to the ones obtained by running a pdf fitting algorithm on the same simulated data , and therefore the fm predictions can be used as a guideline in survey planning .
the advantage of using the fisher matrix as opposed to full pdf fitting is that the former method is much faster , allowing one to very quickly predict the expected results for many different experimental setups .
the paper is organized as follows . in sec .
[ seci ] we review the currently available tools for sed fitting .
[ secii ] describes the relevant fisher information matrix formalism .
our main results for the comparison of uncertainties obtained via the fm and mcmc sed fitting are presented in sec .
[ seciii ] ; we consider several cases of galaxies observed with different signal to noise , star formation histories , dust content , and wavelength coverage . in sec .
[ seciv ] we apply galfish , the fisher matrix algorithm we developed , to real data , using a catalog of several hundred galaxies at redshifts @xmath1 in the great observatories origins deep survey south ( goods - s ) field , and we compare the results to those obtained by using galmc . we provide a worked example of survey optimization in sec .
[ secvi ] contains a summary of our findings and our conclusions .
sed fitting is the procedure of comparing models to the observed galaxy seds ( for reviews , see and references therein ) .
the unknown physical properties of the observed galaxy can be inferred from the known physical properties of the model , once the model that most closely resembles the data ( the best - fit model ) is found .
the best - fit model can be located easily by using @xmath2 minimization techniques , building a grid of models and computing the @xmath2 for each model in the grid .
however , the computation of the uncertainties on sed fitting parameters ( also called _ marginalization _ ) involves solving multi - dimensional integrals , and grid - based techniques are highly inefficient in performing this task . as a result
, markov chain monte carlo techniques have become increasingly popular in the past few years . in mcmc
, the parameter space is explored non - uniformly , sampling more frequently where the models resemble the data more closely . as a result ,
the process is more cpu - efficient , and computing uncertainties amounts to performing sums at the locations sampled by the markov chain . examples of mcmc algorithms for sed fitting include . in @xcite , we presented our own mcmc fitting algorithm , galmc , which is publicly available at http://www.physics.rutgers.edu/@xmath3vacquaviva/web/galmc.html . in this paper
, we use galmc to simulate and fit the data , and to compute the marginalized probability distributions for the sed fitting parameters to be compared with the fm predictions .
let us assume that we are observing galaxies in _
n _ photometric bands , so that the data is a vector containing _
fluxes @xmath4 , each observed with photometric error @xmath5 = @xmath6 . assuming that our models depend on a set of _ m _ parameters @xmath7 , where we include in this list only the parameters we want to measure , the theoretical fluxes in the same photometric bands are denoted as @xmath8 . for the moment , we assume flat priors in the parameters @xmath9 , so that the posterior distribution coincides with the likelihood ; we return to this issue later in this section .
the likelihood associated to each model is = e^- ^2 = e^ _ k = 1^n let @xmath10 be the location of the best - fit model in the parameter space .
around @xmath11 , which is a maximum of the likelihood , we can expand @xmath12 in a taylor series as : = ( ) + _ ij ( _ i - _ i,0 ) _ _ 0 ( _ j,0 - _ j ) + .... [ eq : loglike ] the fisher matrix @xmath13 , which enters the second term in the above equation , is defined as : f_ij = - [ eq : f ] the following relation holds for the marginalized error on parameter @xmath14 in terms of the fisher matrix : [ eq : sigma ] _ _ i .
equation offers the most general way to compute the fisher matrix .
however , in our case this formula can be simplified considerably by using the reasonable assumption that the data obey a gaussian distribution ( in other words , that the likelihood can be written as a gaussian function of the observable quantities ) . in this case , it can be shown ( @xcite ) that f_ij & = & tr [ c^-1 _ i c c^-1 _ j c + & + & c^-1 _
i ^th ( ) _ j ( ^th())^t + & + & c^-1 _ j ^th ( ) _ i ( ^th())^t ] , where @xmath15 is the data covariance matrix , and @xmath16 ; the trace is computed over the length of the data vector . in our case , the photometric errors do not depend on the location of the fiducial model in parameter space , and the first term in the above equation vanishes .
the fisher matrix formalism is able to describe correlations between measurements in different bands as non - diagonal terms in the covariance matrix .
here we make the simplifying assumptions that they are uncorrelated , and the fisher matrix becomes simply f_ij = _ k = 1^n _ j ^th_k ( ) _ i ^th_k ( ) . as is well - known , the fisher matrix does not require the use of any data .
the above formula also allows us to compare the computational advantage of using the fisher matrix to predict the uncertainties on sed fitting parameters , with respect to mcmc methods .
this comparison is only valid for the specific issue of comparing estimated errors before taking the data , since the fisher matrix is not a data fitting technique and , unlike mcmc , does not allow one to find the truly best fitting model and associated uncertainties . in both cases , almost all of the cpu time
is spent running a stellar population synthesis ( sps ) code , in our case the faster version of galaxev @xcite released with galmc , to compute the theoretical model as a function of the sed fitting parameters .
the first derivatives of the model with respect to each parameter can be approximated by the finite differences : _
i ^th_k ( ) = ( ^th_k ( + ) - ^th_k ( - ) ) /(2 _
i ) , [ eq : der ] where @xmath17 is a small increment in the direction of the @xmath18-th parameter . therefore
, each derivative requires computing two theoretical models , and the fisher matrix requires a total of @xmath19 iterations , where @xmath20 is the number of sed fitting parameters .
we have empirically determined in @xcite and @xcite that the necessary number of iterations for galmc to converge is , as a rough rule - of - thumb , @xmath21 , larger by a factor of @xmath22 .
the purpose of this paper is to verify whether ( or under what circumstances ) the @xmath23 " sign in eq .
becomes effectively a @xmath24 " sign .
if the uncertainties from the fisher matrix are systematically much smaller than the ones found by the slower but more accurate parameter estimation methods which are able to reconstruct the form of the pdf , the fisher matrix can not be used as a guideline in planning experiments .
furthemore , because the fisher matrix always assumes a fiducial model , it should only be used in planning experiments and it can not be a substitute for data fitting techniques such as mcmc .
we expect that differences between the uncertainties obtained by fm and pdf fitting can arise from ( at least ) three different issues : - _ a non - gaussian likelihood_. from eq .
, it is clear that if the likelihood is not a gaussian function of the _ parameters _ , the truncation in the taylor series is not exact .
this is different from the aforementioned assumption of a gaussian distribution of the data .
the latter assumption is well justified at high s / n , while the likelihood is strictly a gaussian function of the parameters only if the model fluxes are linear functions of the parameters . for a simple star formation history ( sfh ) , this can be expected to be true for stellar mass , which is an overall normalization of the sed , but not , for example , for age or dust content . - _ a displacement between best - fit model and fiducial model_. the fisher matrix is computed at some fiducial parameter values , chosen to represent a typical target galaxy . however , due to the photometric error , the data will be in general scattered away from the fiducial values of the fluxes , and the best - fit model might not lie at the fiducial location in parameter space , as shown in fig . [
fig : sed ] .
if the shape of the posterior varies significantly with the location in parameter space , the size of the uncertainties predicted by the fm at the fiducial values may not reflect those at the location of the best - fit model ( and as a consequence , those found by pdf fitting ) . in the case of real data ,
this issue can be further worsened by template incompleteness , although this is not a specific drawback of the fm formalism but rather a consequence of assuming a fiducial model .
- _ the impact of priors used in pdf fitting_. in the expression for the fisher matrix , we have implicitly assumed flat priors in the parameters .
while this is not in general necessary ( and , for example , using a uniform prior in @xmath25(age ) or @xmath25(mass ) is straightforward ) , it is not possible to use top - hat priors ( define a range ) in the fm formalism .
this issue is relevant especially for the upper bound in age corresponding to the age of the universe .
such prior can not be taken away from the mcmc run used for pdf fitting , not only because it is an important piece of knowledge that we want to incorporate in the sampling process , but also because theoretical models are only available in a limited age range . however , there are ways to simulate priors of similar effect in the fm , as explained in sec .
[ sec : lm ] . in the next sections we present the results of our careful comparison of the uncertainties predicted by fm and mcmc sed fitting , commenting on each of these issues .
we conduct our analysis on different sets of reference galaxies . the assumed wavelength coverage and limiting magnitudes ( summarized in table [ table : mag ] ) are inspired by fig . 3 of @xcite , which corresponds to the wide " program of the candels galaxy survey @xcite , supplemented by photometry in the spitzer irac 3.6 and 4.5 @xmath26 m bands from goods @xcite
therefore , the results found here represent what is achievable , using public data only , for a galaxy in goods - s .
we do not include any additional calibration uncertainty .
we consider 1 gyr old galaxies at different redshifts ( z = 1 , 2 , and 3 ) with high ( @xmath27 ) and low ( @xmath28 ) stellar mass .
we marginalize over age , total mass m@xmath29 ( the integral of the star formation history over the age of the galaxy , which is the input of sps codes ) , and dust content parametrized by the excess color e(b - v ) . fig .
[ fig : sed ] shows the simulated data ( including a scatter of width equal to the photometric error ) , the fiducial model sed , and the best - fit sed for the high - mass and low - mass reference galaxy at z = 2 , together with the filters transmission curves . .
the mock data points are obtained by adding a gaussian scatter of @xmath30 amplitude equal to the photometric error to the model fluxes ; the best - fit model comes from sed fitting using galmc .
for the low - mass galaxy the displacement between the best - fit and the fiducial model is more pronounced , since the signal to noise ratio is lower and the data have less constraining power .
the filter transmission curves ( multiplied by 0.001 ) are also plotted for reference . ] for all the galaxies described below , we compute the derivatives of the sed ( convolved with the 9 filter curves ) using eq . .
the derivative with respect to the total mass can also be computed analytically since the sed depends linearly on the total mass , and we verified that the derivatives obtained with the two methods agree .
the step size is chosen to be 5% of the reference parameter value , with the exception of dust - free galaxies , where we used values of e(b - v ) = 0.01 and -0.01 ( which is unphysical , and only used in order to correctly evaluate double - sided derivatives ) .
we checked that the step size is unimportant by choosing different values and verifying that the derivatives do not change .
the mcmc analysis in galmc is conducted using a flat prior in @xmath25(age ) and @xmath25(mass ) . in the fisher matrix , we used both a flat prior in age and @xmath25(age ) , and mass and @xmath25(mass ) , finding no or very little ( at the percent level ) dependence on this choice of prior , which wo nt be further discussed . ) .
each row corresponds to a different redshift ( z = 1 , 2 , and 3 ) .
the fm is computed at the fiducial value ( blue thicker line ) and at the best - fit model obtained by galmc when scatter is added to the data ( thin green line ) .
the pdf from galmc is shown for simulated data without ( dashed line ) and with ( dot - dashed line ) scatter .
the estimates from fm and galmc are in good agreement in all cases ; see text for more details . ]
the first set of galaxies we consider are massive galaxies ( @xmath27 ) at the three reference redshifts z = 1 , 2 and 3 , characterized by constant star formation ( csf ) history and a moderate amount of dust ( e(b - v ) = 0.3 ) .
we begin by simulating data with uncertainties corresponding to the magnitude limits in table [ table : mag ] , but not adding scatter to the simulated data points . in this case
, we are testing the first of the caveats " listed in the previous section ; we expect the constraints coming from galmc to be equal to or broader than those predicted by the fm , as a result of a non - gaussian pdf which can not be captured by the fm formalism .
our results are shown in fig .
[ fig : dusty_mas ] .
the agreement between galmc and fm is very good for z = 1 and z = 2 galaxies , the main difference being some non - gaussian tails coming from the mcmc analysis . in the case of the galaxy at z = 3 , the fm is obviously unable to capture the bi - modality of the posterior , and the fm moderately underestimates the uncertainties ( at the 20% level ) .
this behavior is probably a consequence of the lower s / n ratio due to the increase in the luminosity distance at higher redshift .
next , we introduce a gaussian scatter in the mock data , of amplitude equal to the photometric error on each data point , and use galmc to fit the new simulated data .
this setting corresponds to what we can expect with real data .
strictly speaking , the fm is expected to represent the pdf at the location of the best - fit model , so finding a difference in the width of the distribution coming from mcmc , which uses the simulated data , and the fm , which only uses the theoretical models at the fiducial value , is not in itself worrisome . however , if the pdf at the best - fit model is very different from the pdf at the fiducial model , the utility of the fm to predict uncertainties before performing an experiment is diminished , since in that case knowledge of the best - fit model is not available . the results are also shown in fig .
[ fig : dusty_mas ] .
we find a good agreement ( within 20% ) between the width of the distribution obtained by galmc using the data including scatter , and the fm previously computed at the fiducial model values . to verify that the disagreement ( and in particular , the smaller width of the dust distribution reconstructed by galmc at z = 2 ) indeed comes from the displacement of the best fit model with respect to the fiducial one
, we also compute the uncertainties coming from the fm method at the location of the best - fit model . in this case
the uncertainties from the fm are correctly equal to or smaller than the uncertainties coming from galmc . in the subsequent analysis
, we only compare the fm uncertainties evaluated at the fiducial model with the distribution obtained by galmc by fitting simulated data including scatter , since this is the relevant case for survey planning . ) .
mock data include a gaussian scatter .
each row corresponds to a different redshift ( z = 1 , 2 , and 3 ) .
the mcmc predictions include a top - hat prior on age . at redshift z = 1 ( first row ) , the s / n is still good enough that the prior is inessential and the fm predictions match the results of galmc , although the uncertainty in mass is underestimated by the fm by about 50% . at redshift z = 2 and z = 3 ( bottom two rows ) , the prior used in mcmc needs to be incorporated in the fm , which would otherwise overestimate the uncertainties .
when the prior is taken into account , the uncertainties from the two methods are in good agreement for all parameters . ] the next set of galaxies that we consider are similar in sfh and dust content to the ones previously discussed , but have lower stellar mass of @xmath31 , and consequently , lower signal to noise .
we expect sed fitting to be more challenging in this case , since the displacement between best - fit and fiducial model will in general be larger , as illustrated in fig .
[ fig : sed ] , potentially causing a sizable disagreement between the width of the pdf at the fiducial model evaluated by the fm and the width of the real " pdf reconstructed by galmc .
the results are shown in fig .
[ fig : dusty ] . at redshift z = 1 , the fm predictions are similar to the results from galmc , although the uncertainty in the mass is underestimated by the fm by about 50% .
the anticipated difference between fiducial model and best - fit model causes the observed displacement of the peaks in the two curves ; however , the widths of the two distributions , which represent the size of the uncertainties , are very similar . at higher redshift
, though , the fm appears to overestimate the attainable precision on all parameters . in this case , the difference is caused by the top - hat prior on age ( between a minimum value of @xmath32 yr and a maximum value of @xmath33 gyr ) used by galmc . to properly compare the two methods ,
one would need to use the same prior in the fm formalism .
however , it is not possible to include a flat range " in the fm , while it is easy to implement a gaussian prior in any parameter just by adding it to the corresponding diagonal terms of the fisher matrix . therefore , we add a broad gaussian prior of width @xmath34(age / yr ) = 1.6 when computing the fm ; we verified that the exact width of the prior ( between 1.3 and 1.75 ) is not essential .
this simple addition is able to reproduce the main effect of the top - hat prior in age used by galmc , and when the prior is enforced , the uncertainties predicted by the fm or found by galmc are in good agreement . as an empirical prescription for fm calculation , we propose in general to run the fm both with and without this prior .
if no dependence on the prior is found , such as , for example , for the low - mass galaxies at z = 1 , or for all the high - mass galaxies previously discussed ( not shown ) , one can be confident in the fm results .
in general , this will be the case when the tails of the distribution predicted by the fm are far away from the edges of the top - hat priors set in galmc .
on the contrary , if the priors set in galmc intersect the distribution predicted by the fm at high likelihood values ( near the center , rather than near the tails ) , one should keep in mind that these uncertainties might be grossly overestimated , and possibly use a gaussian prior in the fm in order to capture the main effect of the top - hat prior used by mcmc .
galmc also uses a top - hat prior in e(b - v ) between 0 and 1 .
while we did nt find this prior to be of importance in this case , since the posterior distribution intersects these extrema at low values of the likelihood , it might be useful in the general case to test that the fm prediction does not change significantly if a gaussian prior in e(b - v ) is used .
a special case is represented by galaxies with very high ( e(b - v ) @xmath24 1 ) or very low ( e(b - v ) @xmath24 0 ) dust content , which we discuss in detail in the next subsection . ) .
mock data include a gaussian scatter .
each row corresponds to a different redshift ( z = 1 , 2 , and 3 )
. since dust - free galaxies have higher signal to noise in the rest - frame uv - optical than their dusty counterparts , the constraints are significantly better than those from fig .
[ fig : dusty ] , and the prior on age is unimportant .
the slight underestimation of uncertainties from fm ( order of 30% ) in this case might be due to the constraint e(b - v ) @xmath35 0 used by galmc , which can not be reproduced by the fm .
however , this effect is mild and well within the uncertainties observed in other cases . ]
dust - free galaxies present a possible challenge for the fisher matrix vs mcmc comparison , because the prior that forbids negative values of e(b - v ) , used in mcmc , can not be incorporated in the fm formalism .
the latter only allows one to introduce priors centered at the fiducial model , and any gaussian prior centered at e(b - v ) = 0 and wide enough not to introduce a spurious extra weight at low - dust values does not have any effect on the shape of the fisher matrix .
the effect of this difference would be an underestimation of the width of the distribution by the fm .
in fact , all models with e(b - v ) @xmath36 are rejected during mcmc sampling , so that more models with non - zero amounts of dust become part of the markov chain during the random walk , making the posterior distribution ( which is proportional to the density of sampled points ) wider than it would be in absence of this constraint . while we observe this effect in our comparison in fig .
[ fig : nodust ] , the difference in the distribution recovered by galmc and the one predicted by the fm is still moderate ( around 30% ) , even for these low - mass galaxies .
we notice in particular that a large part of the visually striking difference in the dust distribution at z = 2 is attributable to the median value for dust being displaced from the fiducial model by about @xmath37e(b - v ) = 0.1 , while the width of the pdfs from fm and galmc is indeed similar . unlike the case of their dusty counterparts , in dust - free galaxies the rest - frame uv and optical flux is not suppressed , enabling a better determination of age and making the use of a prior in age unnecessary . ) .
mock data include a gaussian scatter .
each row corresponds to a different redshift ( z = 1 , 2 , and 3 ) . for these galaxies the parameter age " , identified with the time from the starburst ,
is 200 myr .
the results obtained by the two methods are in good agreement .
for the z = 3 galaxy we also show the effect of a broad gaussian prior in age ( red dashed - dotted line ) and in e(b - v ) ( green thin solid line ) .
the former has a moderate impact on the prediction for age , while the latter is basically negligible .
the priors have no effect for z = 1 and z = 2 galaxies , as expected ( see discussion in the text ) . ]
to verify whether the star formation history has an impact on the fm vs pdf fitting comparison , we used the two methods to estimate the uncertainties on sed fitting parameters for low - mass starburst galaxies , often referred to as simple stellar populations " ( ssp ) . in this case
, we use the terminology age " slightly inappropriately , since we are basically only sensitive to the time elapsed since the instantaneous burst of star formation happened . in this scenario ,
the rest - frame uv and optical radiation coming from massive stars decreases rapidly with time , since these stars have shorter life spans , and these energetic photons are not replenished by ongoing star formation as in the csf case .
one can not hope to recover the galaxy properties well for a burst that happened 1 gyr before observation ( , at age = 1 gyr ) .
therefore , we take as reference cases galaxies where the time since the starburst is 200 myr .
the results of our comparison , shown in fig . [ fig : ssp ] , confirm that the agreement between the width of the distributions predicted by the fm and computed by galmc is good ( within 20% after the prior on age is properly taken into account for the higher - redshift objects , with the exception of the dust content for the z = 2 galaxy , which is underestimated by a factor two ) . for z = 3 galaxies
, we noticed that the prior at e(b - v ) = 0 used by galmc
cuts " the fm distribution at a high likelihood value .
to possibly reproduce this behavior in the fm formalism , similarly to what we did with age , we introduced an additional wide ( @xmath37e(b - v ) = 0.8 ) gaussian prior in the dust distribution , centered at the fiducial model ( e(b - v ) = 0.3 ) .
we found however that the effect of such prior is negligible ( bottom - right panel of fig .
[ fig : ssp ] ) . as a final test case for our comparison , we considered low - mass galaxies for which no irac coverage is available .
it is well known that the rest - frame near - infrared spectrum is a good tracer of stellar mass , so we expected that determining the galaxy mass would be extremely challenging , and we aimed to verify that the fm formalism would be able to capture this phenomenon .
our results are shown in fig .
[ fig : noirac ] . at redshift z = 1 , the handle offered by the near - ir yjh coverage is still able to guarantee a measurement of stellar mass within a factor of two , and the prior in age has a very moderate effect on the shape of the fm . at higher redshift
, the complete lack of constraints on the sed beyond rest - frame optical wavelengths causes a dramatic increase in the uncertainties in all sed fitting parameters . in these cases , imposing a prior in age in the fm formalism , as discussed in sec .
[ sec : lm ] , becomes important .
after the prior has been incorporated , we find once again good agreement between the prediction of the fm and galmc s results , although , as expected , the parameters are basically unconstrained within their ranges .
the comparison between the results of the fisher matrix approximation and sed fitting has so far only been applied to a set of simulated data . however ,
when the mcmc algorithm is used on real data , we can expect further differences to be introduced by template incompleteness ( the fact that real galaxy spectra are more complicated than the models we used to describe them ) and by the residual effect of priors , as noticed above . to investigate this issue
, we use a sample of 640 galaxies at @xmath38 with spectroscopic redshifts in the goods - s field , which are part of the catalog presented in @xcite .
each sed contains twelve data points , vlt / vimos ( u band ) , hst / acs ( f435w , f606w , f775w , and f850lp bands ) , vlt / isaac ( j , h , and ks bands ) , and the four spitzer / irac channels ( 3.6 , 4.5 , 5.8 , and 8.0@xmath26 m ) .
a @xmath39 flux error in the u , optical and nir bands , and a 10@xmath40 error in the irac bands is added in quadrature to the photometric error to account for calibration uncertainty and template incompleteness .
reveal the heterogeneity of the sample .
the redshift distribution is also shown in the rightmost panel . ] for each of these galaxies , we use galmc to perform mcmc sed fitting and select the objects for which the chains pass all the convergence tests , and the reduced @xmath2 value is less than 10 . in fact , sed fits with very high reduced @xmath2 value indicate a strong template incompleteness problem , and the mcmc uncertainties are likely to be significantly underestimated in these cases .
this procedure selects 485 objects , characterized by a broad range of physical properties ; the distribution of the mean expectation values of sed fitting parameters and the redshifts for the sample are illustrated in fig .
[ fig : params ] .
we proceed to compute the uncertainties on the sed fitting parameters predicted by our fisher matrix algorithm , galfish , for this set of galaxies . in each case
, we use as reference spectrum ( around which the derivatives are computed ) the one of the model characterized by the mean expectation values of the parameters for that galaxy .
the results of our test are shown in fig .
[ fig : fmgalmc ] , which summarizes the differences between the uncertainties found by the mcmc and the ones predicted by the fm , expressed by the quantity @xmath41 for each of the parameters .
( first row ) , signal - to - noise ratio in the sed ( second row ) , and star formation rate ( third row ) .
we do not observe any significant correlation between these properties and the ability of galfish to match the uncertainties of the mcmc analysis.,title="fig : " ] ( first row ) , signal - to - noise ratio in the sed ( second row ) , and star formation rate ( third row ) .
we do not observe any significant correlation between these properties and the ability of galfish to match the uncertainties of the mcmc analysis.,title="fig : " ] ( first row ) , signal - to - noise ratio in the sed ( second row ) , and star formation rate ( third row ) .
we do not observe any significant correlation between these properties and the ability of galfish to match the uncertainties of the mcmc analysis.,title="fig : " ] in the case of age , the average value of the relative difference is 10% , meaning that if the uncertainty on age predicted by fm is 20% , the uncertainty found by mcmc would be @xmath42 .
an useful way to summarize the overall behavior of the test is the fraction of objects for which the absolute value of the relative difference is less than 0.5 or less than 1.0 , which are the 68 and 91% of the sample .
one possible caveat associated to this statistic is that objects for which @xmath43 would still satisfy the condition @xmath44 1.0 , despite the disagreement between the uncertainties predicted by the two methods .
however , such occurrences ( which amount to about 10% of the objects that pass the test ) are much more likely to be due to mcmc underestimating the uncertainties because of template incompleteness , rather than to the failure of the fisher matrix formalism .
in fact , if we alleviate the template incompleteness problem by considering only the subset of seds for which @xmath45 , these occurrences basically disappear .
we also note that the median value of the relative difference is slightly negative ( - 5% ) , while one would generally expect this quantity to be positive , since the fm should give an optimistic evaluation of achievable uncertainties .
however , for this test we have assumed only an extremely broad gaussian prior in age , which basically has no effect on the parameters , while the mcmc uses a top - hat prior on this quantity . therefore , this result is not problematic .
the results for the uncertainties in the stellar mass are similar to those obtained for the age . in this case , the average relative difference @xmath41 is 25% , the fraction of objects for which the absolute value of the relative difference is less than 0.5 or less than 1.0 is 63 and 85@xmath40 respectively , and the median value of the relative difference is again @xmath46 .
the trend is exactly equivalent to what is found for the age , as expected since there is a well - known positive correlation in the increasing age vs increasing mass direction for constant star formation .
finally , the dust content parameterized by e(b - v ) is the parameter for which the fm performs best ; the average relative difference between the uncertainties found by the two methods is 10@xmath40 , the median disagreement is @xmath47 , and the fraction of objects for which the absolute value of the relative difference is less than 0.5 or less than 1.0 is 75@xmath40 and 96@xmath40 respectively .
these statistics are also reported in table [ table : fmgalmc ] .
we also investigate whether the capability of galfish to match the uncertainties obtained via mcmc is correlated with some properties of the target galaxies .
in particular , in fig . [
fig : fmgalmc ] we show the relative differences as a function of the @xmath2 of the fit , of the signal - to - noise ratio of the sed , defined as @xmath48^{0.5}$ ] , and of the star formation rate of the galaxy , but we do not observe any significant trend .
further tests done on the correlations of each parameter with the difference in its own uncertainty similarly do not reveal any recognizable correlation .
in this section we give a practical example of how to optimize observations using galfish .
we assume that spitzer - irac observations at the depths of goods - s are available for our target galaxies ; we fix the depth of these channels to the ones described in @xcite since limited improvements in these measurements will be available in the near future .
our scope is to plan hst observations in the b , v , i , and z channels on the advanced camera for surveys ( acs ) and the y , j and h channels on the wide field camera 3 ( wfc3 ) , with a total of 14 requested orbits .
overhead times are not included in this total time .
we require that at least five out of these seven filters must be used , that an integer number of orbits is spent in each filter , and that a maximum of six orbits are spent in one single filter ; these requirements are merely dictated by common sense and serve to limit the total number of possibilities while preserving the freedom intrinsic in this optimization problem .
each optimization process is valid for a certain type of target galaxy ; in this example , we consider a low mass ( m@xmath49 @xmath50 m@xmath51 ) , 1 gyr old , moderately dusty ( e(b - v ) = 0.2 ) galaxy at @xmath52 , and we run galfish on each of the 9794 possible observational configurations . in fig . [
fig : example ] , we show the predicted fractional uncertainties on age , mass and dust content , as well as their sum , which might be used as the overall quantity to minimize . when this indicator is used , the favored configuration is one where one orbit is allotted to the b , v , and i , and y filters , four orbits are spent in the z filter , six orbits are allotted to the j filter , and the h filter is not used .
the resulting best - case uncertainties achievable on each parameter are @xmath37age / age = 70% or @xmath37age = 0.7 gyr , @xmath37m@xmath49/m@xmath49 = 30% or @xmath53m@xmath49 = 3@xmath54 m@xmath51 , and @xmath37e(b - v)/e(b - v ) = 30% , or @xmath37e(b - v ) = 0.06 .
it is interesting to note that the configurations that lead to the lowest fractional error in each parameter are roughly the same as the ones that give the lowest sum ; in particular , the aforementioned setup is also optimal to minimize the uncertainty on age and mass , and very slightly sub - optimal ( with a difference at the percent level ) to minimize the uncertainty on e(b - v ) . for all parameters ,
we observe a
large scale " modulation which is set by the amount of time spent in the acs b and v filters ; better results are always obtained when one and only one orbit of the total 14 is dedicated to each of these two filters .
the only other constant requirement observed in the minimal - errors region is that the time spent observing with the wfc3 y filter is between one and three orbits .
after these variables have been fixed , several hundred configurations generate similar predicted results ; the 50 configurations generating the lowest error for each parameter are shown in fig .
[ fig : config ] .
in this paper we have compared the uncertainties on sed fitting obtained by using two different methods , the fisher matrix formalism and a mcmc algorithm for pdf fitting , galmc . in the first case ,
the forecasted uncertainties come from a mathematical approximation of the posterior distribution function at some fiducial value of the parameters ; no data are involved . in the second case ,
the data are used to properly reconstruct the pdf by sampling the parameter space through a markov chain .
the latter method is much slower but more accurate ; in particular , the width of the pdf predicted by the fm is in general predicted to be smaller than the real precision attainable on the parameters measured by data fitting techniques such as mcmc .
however , if it can be verified that the uncertainties obtained via the two methods are comparable , the much faster fm becomes a very useful tool to predict the results of a certain experimental setup and to plan astronomical observations . we have compared the pdfs predicted by the fm and fitted by galmc for several sets of reference galaxies , both on simulated data and real data .
the comparison that uses simulated data is the most relevant one for the purpose of comparing the two methods in the context of planning observations , since in any case a fiducial model needs to be chosen , whether the fm or mcmc is used to compute the uncertainties on the sed fitting parameters . on the other hand ,
performing the comparison on real data allows one to test whether the simplifying assumptions used in modeling the stellar populations invalidate the attempt itself to predict the achievable precision in the measurement of parameters . in the case of simulated seds , we started by considering high - mass galaxies ( @xmath55 ) , but most of our analyses have been conducted on low - mass galaxies ( @xmath56 ) , for which more potential sources of disagreement are present as a consequence of the lower signal to noise ratio .
we considered dusty and dust - free galaxies , with constant and instantaneous star formation history , with and without infrared photometric coverage , at redshifts z = 1 , z = 2 , and z = 3 .
we found a remarkably good agreement between the results from the two methods .
of course , the fm is not able to capture features such as non - gaussianity or multi - modality of the posterior , which mcmc can correctly explore .
however , the precision in the sed fitting measurements for the two techniques are within 30% of each other ( with the fm always being an underestimate ) in the vast majority of cases , and within a factor of two in all cases .
priors play an important role in the comparison , in particular for higher redshift galaxies with low s / n .
mcmc uses a top - hat prior in age , ( 6 @xmath57(age / yr ) @xmath58 10.13 ) , bracketed by the minimum age for which templates are trustworthy and by the universe age .
this prior needs to be incorporated in the fm at least approximately , otherwise the fm may grossly overestimate the uncertainties .
we have found that if the fiducial model is far from the extremum values of the age range , a good proxy for the top - hat prior in age is a gaussian prior of width @xmath59(age / yr ) @xmath24 3.5 . as a general prescription , one can always test the predictions with or without the prior in age , but it is easy to verify that the prior only has a non - negligible effect if the prior intersects the distribution predicted by the fm at high likelihood values .
we also investigated the effect of the range imposed in e(b - v ) ( between 0 and 1 in galmc ) , finding this prior to be negligible for all cases considered .
we then proceeded to test the performance of the fm approximation for a set of 485 galaxies at redshifts between 0 and 5 observed in goods - s , with widely spread photometric and stellar population properties . in this case
, we also found that the fm provides an acceptable approximation to the probability distributions found by galmc , with average relative differences between results of the order of @xmath60 , and significant failures ( errors differing by more than a factor of two ) in @xmath3 10% of all cases .
as mentioned earlier , these difference do not necessarily arise from the failure of the fm formalism , but can come from the fact that a real galaxy sed might be more complicated than the model used for the analysis , and the fiducial model used to predict the uncertainties might be significantly different from the observed spectrum . in this analysis , we chose to only marginalize over three sed fitting parameters : age , mass , and dust content .
it is often the case that other parameters are considered in sed fitting , for example , the e - folding time @xmath61 for exponential star formation histories , the metallicity @xmath62 , or the redshift .
as far as @xmath61 is concerned , we showed in @xcite that for exponential sfhs the parameters age and @xmath61 are nearly degenerate , and that different choices of prior on @xmath61 can strongly influence the results of sed fitting , even for those data , which have deeper photometry than the ones considered in this paper .
therefore , @xmath61 is not a suitable parameter for fm analysis .
however , an exponentially declining or decreasing sfh with @xmath63 age is effectively a starburst either at the age of the galaxy or at very recent times , and an exponentially increasing or declining sfh with @xmath64 age is well approximated by a constant sfh , so that the results found here apply to most exponential sfh models .
similarly , the seds depend on metallicity in a very non - linear manner and thus the effect of @xmath62 can not be effectively captured in the fm formalism .
the present analysis assumes that the metallicity of the target galaxies is known , which is a fairly common assumption made in the literature when analyzing broad - band photometry data .
however , if one were to relax this hypothesis , the width of the recovered pdfs could increase .
our analysis in @xcite showed that this effect is mild in the case of @xmath65 and @xmath66 lyman alpha emitters with uv , optical , nir and irac photometry , with probability distributions becoming @xmath3 10 - 20% broader , but of course this effect might be stronger for different stellar populations and data sets and this caveat should be kept in mind . in the case of redshift , a direct comparison with galmc is not really informative , since for real data the redshift is usually determined by using photo - z algorithms .
we have been so far very cautious in using galmc to determine photometric redshifts .
in fact , when the pdf has multiple and widely separated peaks , even mcmc codes might fail to explore the parameter space in a way that respects the basic mcmc property that the density of visited points is proportional to the volume enclosed in the probability distribution function .
as a result , the analysis presented here is strictly valid only if spectroscopic redshifts are available .
however , the fm formalism is able to include redshift as one of the sed fitting parameters , at least in the particular form of a gaussian prior centered at the photometric redshift of the galaxy .
if the uncertainties on the photo-@xmath67 can be expressed in this form , and an educated guess of the width of such prior is available , we can expect the uncertainties from the fm to be reliable . in summary
, we have demonstrated that the fisher matrix formalism is a valuable tool for planning photometric observations .
estimates of the attainable precision on sed fitting parameters are relatively accurate ( within a factor of two at the maximum ) and can be obtained very quickly , allowing one to optimize planned surveys by simulating a variety of experimental setups . just as one example of the many applications , suppose that the scientific goal of an experiment is to achieve a certain precision in the measurement of stellar mass .
the fm formalism allows one to explore what wavelength coverage and depth of observations is needed , and to understand how the aim can be achieved by minimizing costs . in the hope that the formalism cast here is useful to the scientific community , the fisher matrix code presented and used here is publicly available as a php website at http://galfish.physics.rutgers.edu
users can specify which filters they want to use from a list of many common ones , the depth of planned observations , and the characteristics of the target galaxies , and they will immediately obtain the results of the fm analysis .
the fortran code used in the more complex optimization problems will also be available on request upon publication of the paper .
+ we warmly thank the anonymous referee of @xcite for her / his invitation to compare the results obtained by galmc to the predictions from the fisher matrix formalism , tomas dahlen for providing us with the acs - z selected tfit catalog used in sec .
v , and lucia guaita , amir hajian , and licia verde for useful comments on the manuscript draft .
this material is based on work supported by the national science foundation under grants ast-0807570 and ast-1055919 .
support for program number hst - go-12060.57 was provided by nasa through a grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , incorporated , under nasa contract nas5 - 26555 . | the spectral energy distribution ( sed ) of a galaxy contains information on the galaxy s physical properties , and multi - wavelength observations are needed in order to measure these properties via sed fitting . in planning these surveys ,
optimization of the resources is essential .
the fisher matrix formalism can be used to quickly determine the best possible experimental setup to achieve the desired constraints on the sed fitting parameters . however , because it relies on the assumption of a gaussian likelihood function , it is in general less accurate than other slower techniques that reconstruct the probability distribution function ( pdf ) from the direct comparison between models and data .
we compare the uncertainties on sed fitting parameters predicted by the fisher matrix to the ones obtained using the more thorough pdf fitting techniques .
we use both simulated spectra and real data , and consider a large variety of target galaxies differing in redshift , mass , age , star formation history , dust content , and wavelength coverage .
we find that the uncertainties reported by the two methods agree within a factor of two in the vast majority ( @xmath0 ) of cases . if the age determination is uncertain , the top - hat prior in age used in pdf fitting to prevent each galaxy from being older than the universe needs to be incorporated in the fisher matrix , at least approximately , before the two methods can be properly compared .
we conclude that the fisher matrix is a useful tool for astronomical survey design . |
we are interested in numerical solutions of the fully nonlinear elliptic monge - ampre equation @xmath1 on a convex bounded domain @xmath2 of @xmath3 with boundary @xmath4 .
the unknown @xmath5 is a real valued function and @xmath6 are given functions with @xmath7 in the non degenerate case and @xmath8 in the degenerate case .
we will also assume that @xmath9 and @xmath10 in @xmath11 .
starting with @xcite , interest has grown for finite element methods which are able to capture non smooth solutions of second order fully nonlinear equations . for smooth solutions ,
the problem was studied in the context of @xmath0 conforming approximations by b@xmath12hmer @xcite and in the context of lagrange elements by brenner and al @xcite .
non smooth solutions can be handled in the context of the vanishing moment methodology @xcite which is a singular perturbation of . in this paper
we give numerical evidence that @xmath0 conforming approximations of a natural variational formulation of converge for non smooth solutions of the two dimensional problem .
this is achieved by discretizing new iterative methods we introduce .
we establish the convergence of the iterative methods under the assumption that the discrete problem has a convex solution .
we prove that such a solution exists when has a smooth convex solution .
we do not assume that has a smooth solution for our iterative methods to converge .
the existence of a convex solution to the discrete problem in the general case and the convergence of the discretization will be addressed in a subsequent paper .
even with such an existence result , it is still a non trivial task to solve the discrete nonlinear systems in situations where has a non smooth solution .
this paper addresses this issue .
we refer to @xcite for a review of numerical methods for fully nonlinear equations of monge - ampre type .
the main technical innovation of this paper is the proof that in the context of @xmath0 conforming approximations , discrete functions near a strictly convex solution are strictly convex .
this explains why convexity did not need to be imposed explicitly in some previous studies .
newton s method remains the most appropriate iterative method for solving the discrete nonlinear equations when has a smooth convex solution .
we give a new proof of convergence of newton s method in the context of @xmath0 conforming approximations .
the results of this paper extend easily to finite dimensional spaces of piecewise smooth @xmath0 functions provided that the approximation property and inverse estimates below hold .
our results can be described in the general context of discretizations by @xmath0 elements of iterative methods for a general nonlinear elliptic equation @xmath13 . in the case of the monge - ampre equation , @xmath14 .
we first describe the iterative methods at the continuous level .
however , we will not address convergence at the continuous level . as an initial guess we take the solution of the poisson equation @xmath15 in @xmath16 on @xmath4 .
we assume that @xmath17 differentiable and consider the sequence of problems @xmath18 where @xmath19 is a linear operator which can be taken as @xmath20 where @xmath21 is the identity operator or @xmath22 where @xmath23 is the laplace operator and @xmath24 is a parameter .
pseudo transient continuation methods @xcite form a general class of methods for solving nonlinear singular equations . in the case of the monge - ampre equation ,
the method consists in solving the sequence of approximate problems @xmath25 here @xmath26 denotes the matrix of cofactors of the matrix @xmath27 .
given @xmath24 , we consider the sequence of iterates @xmath28 this can be interpreted as an euler discretization of the pseudo time dependent equation @xmath29 or as a laplacian preconditioner of a simple pseudo time marching algorithm , @xcite @xmath30 see also a remark in @xcite . to the author s best knowledge ,
this is the first time the pseudo transient continuation method and the time marching method are used to indicate numerically convergence to viscosity solutions of finite element type methods for the monge - ampre equation .
the methods we propose can be used in the context of different types of discretizations allowing us in particular to treat more easily non - rectangular domains .
the methods can be accelerated with fast poisson solvers and multigrid methods .
this latter property is even more striking for the time marching method as its implementation requires only having access to a multigrid poisson solver .
although the theory of the monge - ampre equation has concentrated on convex solutions , one can equally focus on concave solutions .
we found out that is better able to capture concave solutions .
it is easy to implement , requiring only a poisson solver .
for example one can capture weak solutions of the monge - ampere equation by simply discretizing with the standard lagrange finite elements .
the time marching method can also be applied to fully nonlinear equations such as the pucci equation where @xmath17 is not differentiable . in summary
the pseudo transient continuation methods are better for smooth solutions and singular solutions on a coarse mesh .
otherwise the method of choice is the time marching method .
we organize the paper as follows : in the second section we introduce some notation and prove the key result that discrete functions near a strictly convex solution are strictly convex .
we introduce the natural variational formulation of and state an existence and uniqueness result for the discrete problem .
the result can be inferred from the ones in @xcite .
we give new proofs in section [ existence ] . as a corollary the discrete variational problem has a convex solution when has a smooth convex solution .
we study the pseudo transient continuation methods in section [ continuation ] .
a special case is newton s method for which we prove a quadratic convergence rate .
the time marching methods are studied in section [ time ] .
the last section is devoted to numerical results .
we give a brief description of the spline element method which is used for the computations and offer heuristics about why our methods appear to preserve convexity .
we use the standard notation for the sobolev spaces @xmath31 with norms @xmath32 and semi - norm @xmath33 . in particular , @xmath34 and in this case , the norm and semi - norms will be denoted respectively by @xmath35 and @xmath36 . for a vector field @xmath37 with values in @xmath38 , we set @xmath39 and a similar notation for @xmath40 . in the case
@xmath41 , we set @xmath42 with a similar notation for @xmath43 .
for matrix valued fields , the above notation is extended canonically .
we make the usual convention of denoting constants by @xmath44 but will occasionally index some constants . for constants which depend on the mesh size
, we may use @xmath45 or @xmath46 .
we make the assumption that the boundary of @xmath2 is polygonal and that the triangulation @xmath47 is shape regular in the sense that there is a constant @xmath48 such that for any triangle @xmath49 , @xmath50 , where @xmath51 denotes the diameter of @xmath49 and @xmath52 the radius of the largest ball contained in @xmath49 .
we also require the triangulation to be quasi - uniform in the sense that @xmath53 is bounded where @xmath54 and @xmath55 are the maximum and minimum respectively of @xmath56 .
we define @xmath57 where @xmath58 denotes the space of polynomials of degree less than or equal to @xmath59 . in _ two dimensions _
, it is known that , @xcite , for @xmath60 and @xmath61 , there exists a linear quasi - interpolation operator @xmath62 mapping @xmath63 into the spline space @xmath64 and a constant @xmath44 such that if @xmath65 is in the sobolev space @xmath66 @xmath67 for @xmath68 . if @xmath2 is convex , the constant @xmath44 depends only on @xmath69 and on the smallest angle @xmath70 in @xmath47 . in the nonconvex case ,
@xmath44 depends only on the lipschitz constant associated with the boundary of @xmath2 .
it is also known c.f .
@xcite that the full approximation property for spline spaces holds on special triangulations for certain values of @xmath59 . in _ three dimensions
, holds in general for @xmath71 , c.f .
@xcite . note that , by , @xmath72 for all @xmath73 .
we assume that the following inverse inequality which holds for @xmath0 finite element spaces , c.f .
theorem 4.5.11 of @xcite , also holds for the spline spaces @xmath74 and for @xmath75 .
the local estimates may be viewed as a consequence of the assumption of uniform triangulation and of markov inequality , @xcite p.
2 . passing from local estimates to global estimates can be done as in @xcite .
we first recall the divergence form of the determinant and the expression of its frchet derivative . for two @xmath76 matrices @xmath77 ,
we recall the frobenius product @xmath78 .
[ det - lem ] we have @xmath79 and for @xmath80 we have @xmath81 for @xmath82 sufficiently smooth
. note that for any @xmath76 matrix @xmath27 , @xmath83 , where @xmath26 is the matrix of cofactors of @xmath27 .
this follows from the row expansion definition of the determinant . for any sufficiently smooth matrix field @xmath27 and vector field @xmath65 , @xmath84 . here
the divergence of a matrix field is the divergence operator applied row - wise . if we put @xmath85 , then @xmath86 and @xmath87 but @xmath88 , c.f .
for example @xcite p. 440 .
hence since @xmath89 and @xmath90 are symmetric matrices follows .
the assertion about the frchet derivative of @xmath17 follows from the definition of the determinant as a multilinear map ( e.g ) and the definition of matrix of cofactors .
see also @xcite p. 440
. using the divergence form of the determinant and integration by parts , one obtains the variational formulation of given by : find @xmath91 , @xmath92 on @xmath4 such that @xmath93 we show that for @xmath91 , is well defined .
_ case @xmath94_. for @xmath94 , each entry of @xmath90 consists of a second derivative @xmath95 . by h@xmath12lder s inequality , @xmath96 next for @xmath97 and by sobolev embedding , i.e. the embedding of @xmath98 in @xmath99 for @xmath100 when
@xmath94 , the right hand side above is bounded by @xmath101 @xmath102 .
_ case @xmath103_. for @xmath103 , each entry of @xmath90 involves the product of two second order derivatives .
we have by h@xmath12lder s inequality and sobolev embedding , i.e. the embedding of @xmath98 in @xmath99 for @xmath104 when @xmath103 , @xmath105 we conclude that for @xmath103 , @xmath106 in summary for @xmath107 , we may write @xmath108 put @xmath109 and @xmath110 .
note that @xmath111 given by satisfies @xmath112 .
let @xmath113 and furthermore let @xmath114 be the interpolant in @xmath111 of a smooth extension of @xmath10 .
we have the following conforming discretization of : find @xmath115 , @xmath116 on @xmath4 such that @xmath117 we now present a number of preliminary results .
we first state a result on the well posedness of .
the @xmath118 error estimate can be inferred from @xcite .
we give a new proof in section [ existence ] .
[ errorest ] let @xmath119 and assume that @xmath120 is a strictly convex function , that @xmath2 is convex with a polygonal boundary and that the spaces @xmath111 have the optimal approximation property and satisfy the inverse estimates .
then the problem has a unique solution @xmath121 for @xmath54 sufficiently small and we have the error estimates latexmath:[\ ] ] the ellipticity assures uniqueness of the component @xmath446 and the saddle point problems are solved by a version of the augmented lagrangian algorithm @xmath447 the convergence properties of the iterative method were given in @xcite .
extensive implementation details can be found in @xcite . for @xmath94 ,
the computational domain is the unit square @xmath448 ^ 2 $ ] which is first divided into squares of side length @xmath54 .
then each square is divided into two triangles by the diagonal with negative slope . for @xmath103 ,
the initial tetrahedral partition @xmath449 consists in six tetrahedra .
each tetrahedron is then uniformly refined into 8 subtetrahedra forming @xmath450 . in the tables
, @xmath451 denotes the number of iterations .
we refer to @xcite for implementation details of the method .
all numerical experiments are with the versions of the iterative methods with laplacian preconditioner . in general we did not try to choose the value of @xmath325 that would give the smallest number of iterations except in tables [ tab - comp1 ] and [ tab - comp2 ] where we compare the performance of the two methods .
we use some standard test cases for numerical evidence for convergence to non smooth solutions of the elliptic monge - ampere equation .
test 1 : @xmath452 so that @xmath453 and @xmath454 on @xmath4 .
test 2 : @xmath455 so that @xmath456 and @xmath457 on @xmath4 .
barring roundoff errors , the methods introduced in this paper capture smooth solutions . for the two dimensional test function , test 1 , we give numerical results for successive refinements and for the three dimensional test function , we give numerical results for increasing values of the degree @xmath59 on two successive refinements .
.time marching method for test 1 , @xmath458 , @xmath459 [ cols="^,^,^,^,^,^,^,^",options="header " , ] the time listed is in seconds and obtained on an imac running mac os 10.6.8 with a 2.4 ghz intel core 2 duo and 4 gb of sdram memory .
while for small values of @xmath54 the time marching method appears to take significantly more time , it is also significantly more accurate .
for @xmath460 the time took by the two methods is almost the same with the time marching method giving a more accurate solution .
next we consider a non square domain .
test 5 : we consider the unit circle discretized with a delanauy triangulation with 824 triangles and @xmath461 which vanishes on the boundary , figure [ fig11 ] . on a non square domain with pseudo transient @xmath462 , height=170 ]
we conclude this section with a test problem for a degenerate monge - ampre equation test 6 : @xmath463 and @xmath464 .
the graph of the function , figure [ fig2 ] is singular along the line @xmath465 .
the approximations been @xmath0 do appear to capture the singularity but not the convexity of the solution .
somewhat better results are obtained with another iterative method discussed in @xcite .
when the time marching method is discretized by the standard finite difference method the singularity is captured correctly .
we wish to discuss these results in separate works . and @xmath463 with time marching @xmath466 , title="fig:",height=170 ] and @xmath463 with time marching @xmath466 , title="fig:",height=170 ] when has a smooth strictly convex solution , theorem [ errorest ] establishes that the approximate solution is automatically convex .
the numerical experiments indicate that in the non smooth case , discrete solutions are also convex .
the result can be easily explained at the continuous level ( for a smooth solution ) .
assume that @xmath467 and that the sequence @xmath250 defined by @xmath468 has been shown to converge to @xmath5 in the h@xmath12lder space @xmath469 for some @xmath470 in @xmath471 . from the arithmetic - geometric inequality , we have @xmath472 by the continuity of the eigenvalues , , @xmath473 is bounded in a neighborhood of @xmath5 in which all @xmath250 belong for @xmath474 large enough .
choose @xmath325 such that @xmath475 for all @xmath474 and note that the right hand of is equal to @xmath476 . by the assumption on @xmath325 , we get @xmath477 . in the limit , we obtain @xmath478 . since @xmath479 by assumption , we get @xmath480 . as for
the time marching method @xmath481 assume now again that @xmath482 and that the sequence @xmath250 has been shown to converge to @xmath5 in @xmath469 for some @xmath470 in @xmath471 .
choose @xmath325 such that @xmath483 .
we have @xmath484 .
it follows from the arithmetic - geometric inequality that @xmath485 and so @xmath486 and it follows that the time marching method also preserves the positivity of the laplacian . in two dimensions @xmath480 and @xmath487 imply that @xmath89 is positive definite .
the author acknowledges discussions with f. celiker , b. cockburn , w. gangbo , r. glowinski , m.j .
lai , r. nochetto , a. oberman and a. regev .
the author was supported in part by nsf grant dms-0811052 and the sloan foundation .
this research was supported in part by the institute for mathematics and its applications with funds provided by the national science foundation .
awanou , g. , lai , m.j . , wenston , p. : the multivariate spline method for scattered data fitting and numerical solution of partial differential equations . in : wavelets and splines : athens 2005 , mod .
methods math . , pp .
nashboro press , brentwood , tn ( 2006 ) davydov , o. , saeed , a. : stable splitting of bivariate splines spaces by bernstein - bzier methods . in : curves and surfaces ,
_ lecture notes in comput .
. 6920 , pp . 220235 .
springer , heidelberg ( 2012 ) dyer , b.w . ,
hong , d. : algorithm for optimal triangulations in scattered data representation and implementation .
j. comput .
appl . * 5*(1 ) , 2543 ( 2003 ) .
approximation theory and wavelets ( austin , tx , 1999 ) farag , i. , kartson , j. : numerical solution of nonlinear elliptic problems via preconditioning operators : theory and applications , _ advances in computation : theory and practice _ , vol .
nova science publishers inc . , hauppauge , ny ( 2002 ) | we present two numerical methods for the fully nonlinear elliptic monge - ampre equation .
the first is a pseudo transient continuation method and the second is a pure pseudo time marching method .
the methods are proved to converge to a convex solution of a natural discrete variational formulation with @xmath0 conforming approximations .
the assumption of existence of a convex solution to the discrete problem is proven for smooth solutions of the continuous problem and supported by numerical evidence for non smooth solutions . |
the large - scale features of the observed universe are strikingly consistent with the theoretical predictions based on the cold dark matter model
. the combined analyses of the recent data from the observations of cosmic microwave background ( cmb ) , galaxy power spectrum and type ia supernovae ( e.g. , * ? ? ?
* and references therein ) have been capable of measuring the key cosmological parameters that characterize the cdm model with surprisingly high precision . this has opened an era of precision cosmology , echoing the triumph of the cdm model .
nevertheless , the status of the cdm model as the standard paradigm has been shaking currently in both observational and theoretical perspectives .
observations have reported several mismatches between the predictions of the cdm model and the real phenomena on galactic and subgalactic scales .
for instance , the abundance of galactic satellites , the slope of the inner core of the dark halo density profiles , and the degree of void emptiness have exhibited apparent conflicts between theory and observation @xcite .
very recently , @xcite measured the cross - correlations between different galaxy s observables and showed that only a single parameter suffices to explain the complex structures of galaxies .
this phenomenon is inconsistent with the cdm picture where the galaxy formation is driven by hierarchical merging process .
although it is still inconclusive whether the reported mismatches really indicate the failure of the cdm model or they are just by - products of the selection effects and/or limitations of the observational techniques , it has certainly led to an emergence of alternative models .
a theoretical challenge comes from the fact that the extension of the standard particle physics favors the warm dark matter ( wdm ) rather than cdm .
the wdm differs from cdm in a respect that the wdm particles have non - negligible free streaming scale , @xmath3 . on scales larger than @xmath3 , wdm behave just like cdm . on the other hand , on scales less than @xmath3 , wdm do not cluster but free stream due to their large velocity dispersions unlike cdm which cluster on all scales .
it is this property that enables the wdm model to overcome the observational mismatches faced by the cdm model .
the list of realistic wdm candidates that have so far proposed includes gravitinos ( * ? ? ?
* and references therein ) , axions @xcite , and sterile neutrinos @xcite , each of which has distinct characteristic mass range .
by constraining the wdm particle mass , it may be possible to rule out a certain wdm candidate .
in fact there have been already such attempts : for instance , @xcite constrained the particle mass of wdm by using the ly@xmath4 forest power spectrum and suggested that the sterile neutrinos be ruled out ( see also * ? ? ?
* ) in a wdm universe , the objects that will first condense out in the initial density inhomogeneities are not zero dimensional clumps but one dimensional filaments .
@xcite recently studied the formation and properties of wdm filaments with high - resolution hydrodynamic simulations , and proposed that the early - universe filaments would seed the supermassive black holes that power the quasars at @xmath5 through their eventual collapse along the longest axes .
if this scenario is true , then , the following two issues should be related to each other : how luminous and abundant the first generation high-@xmath1 quasars are ; and how massive and abundant the early - universe wdm filaments are . in statistical terms
, the luminosity function of the high-@xmath1 quasars should be related to the mass function of the early - universe filaments .
given that the mass function of the early - universe filaments depends on the particle mass of wdm ( see 2 ) , it implies the possibility of using the high-@xmath1 quasar luminosity function as a new wdm constraint . the idealistic way to study the mass function of early - universe filaments and
its dependence on the wdm particle mass is to use the hydrodynamics simulations .
however , given the computational costs as well as the current resolution - limit of hydrodynamic simulations , it is adequate and quite necessary to consider an analytic approach as the first guideline . here , we evaluate analytically the abundance of early - universe filaments and constrain the wdm particle mass by fitting its slope in the wdm particle free streaming scale to that of the high-@xmath1 quasar luminosity function @xcite determined from the sloan digital sky survey data release 3 ( sdss ) @xcite . for the key cosmological parameters other than the wdm particle mass , we assume a wmap5 cosmology @xcite throughout this paper . the plan of this paper is as follows . in
2 , an analytic model for the abundance of early universe filaments is presented . in 3 , a constraint on the particle mass of wdm is provided by fitting the analytic model to the observational data from sdss . in 4 , the achievements as well as the caveats of our work are discussed .
the zeldovich approximation predicts generically and in the most simple way the formation of early - universe filaments @xcite . according to this model
, the mass density @xmath6 of a given region can be expressed in terms of the lagrangian quantities as @xmath7[1-d(z)\lambda_2][1-d(z)\lambda_3 ] } , \label{eqn : rho}\ ] ] where @xmath8 is the mean background density , @xmath9 ( in a decreasing order ) are the three eigenvalues of the linear deformation tensor , and @xmath10 is a linear growth factor .
equation ( [ eqn : rho ] ) predicts that @xmath6 will diverge at @xmath11 .
that is , the first collapse will occur when the largest eigenvalue reaches a threshold value , @xmath12 .
the sign of the other two eigenvalues , @xmath13 and @xmath14 , at the moment of the first collapse will determine the dimension of a collapsed object .
a one - dimensional filament will form if @xmath15 and @xmath16 , while the formation of a two - dimensional sheet will occur for the case @xmath17 and @xmath16 .
since we are interested in the formation of `` filaments '' that are shown to be most abundant in the early universe filled by wdm @xcite , we consider only the filament condition , @xmath11 , @xmath15 and @xmath16 .
the fractional volume @xmath18 occupied by those lagrangian regions which will condense out early - universe filaments at @xmath1 on mass scale @xmath19 can be expressed as @xmath20,\ ] ] where @xmath21 and @xmath22 is the rms density fluctuation related to the dimensionless linear wdm power spectrum @xmath23 as @xmath24 where @xmath25 represents a sharp k - space filter on mass scale @xmath19 .
we adopt the approximation formula of@xcite , @xmath26 where @xmath27 is the cdm linear power spectrum .
the functional form of the wdm transfer function , @xmath28 is given as @xmath29^{-5/\gamma}\ ] ] where @xmath30 @xmath31 , @xmath2 is the wdm particle mass and @xmath32 is the number of degrees of freedom that equals @xmath33 for fermions . the fractional volume in equation ( [ eqn : fm ] ) can be obtained by integrating the joint probability density distribution of the three eigenvalues that was first derived by @xcite as @xmath34 where @xmath35 and @xmath36 . in the spirit of the press - schechter theory (
* hereafter ps ) , we evaluate the number density of early - universe filaments collapsed at mass scale @xmath19 as @xmath37 where the redshift for the formation of the early - universe filaments is set at @xmath38 , in accordance with @xcite . in the original ps theory which dealt with the mass function of the clumps , equation ( [ eqn : nf ] )
had to be multiplied by an additional normalization constant to account for the occurrence of the cloud - in - clouds @xcite .
but , for our case the cloud - in - clouds will not occur since the wdm particles just free - stream in the early - universe filaments forming no substructures .
henceforth , we do not have to account for the cloud - in - cloud problem in deriving the mass function of early universe filaments . the cumulative mass function @xmath39 of early universe filaments
can be obtained by integrating equation ( [ eqn : nf ] ) as @xmath40 . since
our focus is not on the amplitude of the mass function but mainly on its slope , we rescale @xmath39 to have unity as @xmath19 goes zero .
figure [ fig : nf ] plots the rescaled cumulative mass function @xmath41 at @xmath38 versus @xmath42 for the case of @xmath43kev ( open circles ) and @xmath44kev ( open squares ) . here
@xmath45 corresponds to the free streaming mass scale of a given wdm particle .
as can be seen , @xmath39 begins to decrease as @xmath19 becomes larger than @xmath45 .
the rate of its decrease at @xmath46 depends sensitively on the value of the wdm particle mass @xmath2 : it decreases more rapidly with @xmath19 for the case of smaller value of @xmath2 .
it is because the less massive wdm particles have smaller free - stream scales . to quantify the rate of the decrease of @xmath39 at @xmath46
, we fit @xmath47 to a power - law @xmath48 ( i.e. , @xmath49 ) in the mass range where @xmath50 . in the higher mass section where @xmath51 drops exponentially with @xmath19 , we expect basically no early universe filaments . as demonstrated by @xcite , only those early - universe filaments with mass comparable in order of magnitude to the wdm free - streaming scale can form which are in turn related to the first - generation supermassive black holes that power the observed high-@xmath1 luminous quasars . the higher mass section where @xmath52 correspond to the scales order - of - magnitude larger than the free streaming scale of a given wdm particles .
the solid lines in fig .
[ fig : nf ] represent the power - law fitting of the cumulative mass function of early universe filaments .
as can be seen , the slope @xmath53 indeed depends sensitively on the value of @xmath2 .
the discovery of a supermassive black hole with mass @xmath54 in quasar sdss j1148 + 5251 at @xmath55 @xcite has raised a crucial question of what seeded such a supermassive black hole at that early epoch ( * ? ? ? * and references therein ) . as introduced in 1 ,
@xcite have studied the formation of early - universe filaments using high - resolution hydrodynamic simulations for a wdm cosmology and suggested that the ultimate collapse of the wdm filaments along the longest axes should induce vigorous collisions between gas clouds and stars due to their high densities , which would in turn seed the formation of supermassive black holes that power quasars at @xmath5 . in subsequent evolution , the black holes would grow active by accreting baryonic gases and dark matter particles that constituted their parent filaments . in this picture , the number density of active black holes at high redshifts would be related to that of early universe filaments .
the luminosity function @xmath56 of high-@xmath1 quasars at @xmath57 was determined by @xcite , who showed that @xmath56 can be well approximated as a power law with high - end slope in the range of @xmath58 .
if these high-@xmath1 quasars are indeed powered by the first - generation supermassive black holes formed through the ultimate collapse of the wdm filaments , the high - end slope of the high-@xmath1 quasar luminosity function should be consistent with the power - law slope of the mass function of early - universe filaments determined in 2 .
varying the value of @xmath2 , we determine the power - law slopes of the mass function of early - universe filaments @xmath53 , which are shown in figure [ fig : slope ] .
the @xmath59 and @xmath60 range of the high - end slope of the high-@xmath1 quasar luminosity function @xmath56 measured from sdss dr3 are shown as horizontal dashed and dotted lines , respectively .
the comparison indicates that the two slopes agree with each other when @xmath2 is a few kev .
motivated by the recent heuristic work of @xcite , we have constructed an analytic model for the mass function of the early - universe filaments in a wdm universe .
then we test the possibility of using it as a new wdm constraint . as a feasibility study
, we compare the high - end slope of the mass function of the early - universe filaments to that of the high-@xmath1 quasar luminosity function determined from sdss dr3 and find a new wdm constraint , @xmath61 a few kev .
this preliminary result looks consistent with the super - symmetric picture @xcite .
the advantage of the mass function of the early - universe filament as a wdm constraint lies in the fact that the model is purely analytical , free of any fitting or nuisance parameter .
it depends only on the wdm particle mass other than the key cosmological parameters .
in addition , when the analytic model is compared to the observational result , it is not required to account for any complicated baryon physics since we consider only the _ slope _ of the mass function . yet
, there are a couple of simplified assumptions on which our model is based .
first , the redshifts of the quasars should be higher than @xmath5 for a more fair comparison . in this work
, we just used the redshift range @xmath57 , assuming that the quasars at @xmath57 correspond to the first generation . to be more realistic
, it will be required to determine the luminosity function of the quasars at @xmath5 , but there are only small number of quasars have so far been observed at that high redshift .
for example , @xcite determined the supermassive black hole mass function at @xmath62 using a tens of quasars and found the high - end slope in a much larger range of @xmath63 , which obviously suffers from large uncertainty due to the small number statistics .
anyway , we would like to mention clearly that the true high - end slope of the high-@xmath1 quasar luminosity function is currently significantly more uncertain than the data used here .
second , the analytic model for the mass function of early universe filaments has to be refined .
our model is the simplest approximation based on the zeldovich approximation and the press - schechter theory .
the validity of our model has to be tested numerically before using it in practice .
third , the evolution of the early - universe filaments has not been taken into account properly for the evaluation of the mass function . in reality , before their eventual collapse at @xmath5 , the early - universe filaments may undergo fragmentation due to the tidal effect from the surrounding matter . to account for the evolutionary effects , it will be anyway required to refine the model with the help of the hydrodynamic simulations .
our future work is in this direction . as a final conclusion ,
our model and the preliminary result from our feasibility study has provided a proof of concept that the high-@xmath1 quasar luminosity function will be in principle a useful wdm probe .
we thank an anonymous referee for helpful comments .
this work is financially supported by the korea science and engineering foundation ( kosef ) grant funded by the korean government ( most , no .
r01 - 2007 - 000 - 10246 - 0 ) .
battye , r. a. , & shellard , e. p. s. 1994 , , 73 , 2954 bertone , g. , hooper , d. , & silk , j. 2005 , , 405 , 279 bode , p. , ostriker , j. p. , & turok , n. 2001 , , 556 , 93 bond , j. r. , cole , s. , efstathious , g. , & kaiser , n. 1991 , , 379 , 440 christlieb , n. , et al .
2002 , , 419 , 904 coles , p. , melott , a. l. , & shandarin , s. f. 1993 , , 260 , 765 croft , r. a. c. , weinberg , d. h. , katz , n. , & hernquist , l. 1998 , , 495 , 44 dalla bont , e. , ferrarese , l. , corsini , e. m. , miralda - escud , j. , coccato , l. , sarzi , m. , pizzella , a. , & beifiori , a. 2009 , , 690 , 537 disney , m. j. , romano , j. d. , garcia - appadoo , d. a. , west , a. a. , dalcanton , j. j. , & cortese , l. 2008 , , 455 , 1082 dodelson , s. , & widrow , l. m. 1994 , , 72 , 17 doroshkevich , a. g. 1970 , astrofizika , 6 , 581 dunkley , j. , et al .
2009 , , 180 , 306 fan , x. , et al .
2001 , , 121 , 54 frebel , a. , et al .
2005 , , 434 , 871 gao , l. , & theuns , t. 2007 , science , 317 , 1527 gtz , m. , & sommer - larsen , j. 2002 , , 281 , 415 hu , j. , shen , y. , lou , y .- q .
, & zhang , s. 2006 , , 365 , 345 klypin , a. , kravtsov , a. v. , valenzuela , o. , & prada , f. 1999 , , 522 , 82 lee , j. & shandarin , s. f. 1998 , apj , 500 , 14 moore , b. , ghigna , s. , governato , f. , lake , g. , quinn , t. , stadel , j. , & tozzi , p. 1999
, , 524 , l19 munyaneza , f. , & biermann , p. l. 2005 , , 436 , 805 narayanan , v. k. , spergel , d. n. , dav , r. , & ma , c .- p .
2000 , , 543 , l103 peebles , p. j. e. 2001 , , 557 , 495 podsiadlowski , ph . , rappaport , s. & hau , z. 2003 , , 341 , 385 press , w. h. , & schechter , p. 1974
, , 187 , 425 seljak , u. , makarov , a. , mcdonald , p. , & trac , h. 2006 , , 97 , 191303 trump , j. r. , et al .
2006 , , 165 , 1 vestergaard , m. , fan , x. , tremonti , c. a. , osmer , p. s. , richards , g. t. 2008 , , 674 , l1 viel , m. , lesgourgues , j. , haehnelt , m. g. , matarrese , s. , & riotto , a. 2006 , , 97 , 071301 willott , c. j. , mclure , r. j. , & jarvis , m. j. 2003 , , 587 , l15 yoshida , n. , sokasian , a. , hernquist , l. , & springel , v. 2003 , , 591 , l1 zeldovich , y. b. 1970 , , 5 , 84 | in a warm dark matter ( wdm ) cosmology , the first objects to form at @xmath0 are one dimensional filaments with mean length on the order of the wdm free - streaming scale .
gao and theuns recently claimed by using high - resolution hydrodynamic simulations that the eventual collapse of these wdm filaments along their longest axes may seed the supermassive black holes that power high-@xmath1 quasars . in this picture , it is supposed that the high-@xmath1 quasar luminosity function should reflect how abundant the wdm filaments are in the early universe .
we derive analytically the mass function of early - universe filaments with the help of the zeldovich approximation .
then , we determine the rate of its decrease in the mass section corresponding to the free streaming scale of a wdm particle of mass @xmath2 . adjusting the value of @xmath2
, we fit the slope of the analytic model to that of the high-@xmath1 quasar luminosity function measured from the sloan digital sky survey dr3 . a new wdm constraint from this feasibility study
is found to be consistent with the lightest super - symmetric partner . |
this section is devoted to study the efficiency of the decision tree algorithm described in the main text .
some results in this section are analytical and some are numerical . in all our numerical investigations (
unless explicitly stated otherwise ) we used the decision tree of the main text ( for two qubits ) and in cases when going through the whole tree did not reveal entanglement we augmented it with additional measurements of those correlations which were not performed until that moment .
the order of the additional measurements also results from the correlation complementarity ( anti - commutation relations ) @xcite . with every remaining measurement we associate the `` priority '' parameter @xmath110 that depends on the measured correlation tensor elements of the decision tree in the following way
@xmath111 according to the correlation complementarity there is a bigger chance that this correlation is significant if the value of the corresponding parameter is small .
therefore , the correlations @xmath112 with lower values of @xmath113 are measured first .
let us consider the following example .
the measured correlations of the decision tree are as follows : @xmath114 , @xmath115 , and @xmath116
. therefore , @xmath117 , @xmath118 , and @xmath119 .
accordingly the order of the remaining measurements is as follows : first measure @xmath52 , then @xmath120 , next @xmath121 and @xmath122 .
consider a family of states @xmath123 where @xmath124 is the bell singlet state , @xmath125 describes the completely mixed state of white noise , and @xmath126 is a probability .
its correlation tensor , written in the same coordinate system for alice and bob , is diagonal with entries @xmath127 , arising from the contribution of the entangled state .
the states ( [ werner_state ] ) are entangled if and only if @xmath128 , whereas the decision tree reveals that these states are entangled for @xmath129 .
this is because only three elements contribute to the criterion .
note that the value of the decision parameter is irrelevant here .
furthermore , a random choice of local coordinate directions does not help .
although more steps would be involved in the decision tree the sum of squared correlations is invariant under local unitary operations and therefore only for @xmath130 entanglement is detected for the werner state . an exemplary class of density operators for which the decision tree detects all entangled states
is provided by : @xmath131 where entanglement is mixed with colored noise @xmath132 bringing anti - correlations along local @xmath46 axes .
this state has the following non - vanishing elements of its correlation tensor @xmath133 and @xmath134 .
therefore , the decision tree allows detection of entanglement for this class of states in two steps .
note that the state is entangled already for an infinitesimal admixture of the bell singlet state as can be shown for a random choice of local coordinate systems .
[ si_fig_random_mixed ] shows how the efficiency of the algorithm grows with the purity of tested states .
the efficiency is measured by the fraction of detected entangled states obtained from extensive monte carlo sampling in the two qubit state space . for nine steps
the algorithm detects all the pure states .
this is expected because eq .
( 1 ) of the main text is a necessary and sufficient condition for the detection of entanglement of pure states .
the principle behind the decision tree is the correlation complementarity @xcite .
correlation complementarity states that for a set of dichotomic anti - commuting operators @xmath135 the following trade - off relation is satisfied by all physical states : @xmath136 where @xmath137 is the average value of observable @xmath138 and so on . therefore , if one of the average values is maximal , @xmath139 , the other anti - commuting observables have vanishing averages .
this motivates taking only sets of commuting operators as different branches of the decision tree .
a numerical simulation reveals that the correlation measurement along local bloch vectors gives correlations close to the maximal correlations of a pure multi - qubit state in more than @xmath53 of the cases .
therefore , these local directions give an excellent starting point of the decision tree .
exemplarily , the branch begins with @xmath97 assuming that this correlation is big . if the local bloch vectors indicate correlation along a different directions it is advantageous to correspondingly change the elements of the decision tree ( see main text ) . | we introduce an experimental procedure for the detection of quantum entanglement of an unknown quantum state with a small number of measurements .
the method requires neither a priori knowledge of the state nor a shared reference frame between the observers and can thus be regarded as a perfectly state independent entanglement witness .
the scheme starts with local measurements , possibly supplemented with suitable filtering , which essentially establishes the schmidt decomposition for pure states .
alternatively we develop a decision tree which reveals entanglement within few steps .
these methods are illustrated and verified experimentally for various entangled states of two and three qubits . _
introduction._entanglement is the distinguishing feature of quantum mechanics and it is the most important resource for quantum information processing @xcite . for any experiment
it is thus of utmost importance to easily reveal entanglement , best with as little effort as possible .
common methods suffer from disadvantages .
on the one hand , employing the peres - horodecki criterion @xcite or evaluating entanglement measures , one can identify entanglement in arbitrary states , however , it requires full state tomography . on the other hand , various entanglement witnesses @xcite can be determined with much fewer measurements but give conclusive answers only if the state under investigation is close to the witness - state , i.e. , they require a priori knowledge .
recently , it has been shown that the existence of entanglement can be inferred from analyzing correlations between the measurement results on the subsystems of a quantum state .
only if the state is entangled , the properly weighted sum of correlations will overcome characteristic thresholds @xcite .
here we further develop this approach to obtain a simple and practical method to detect entanglement of all pure states and some mixed states by measuring only a small number of correlations .
since the method is adaptive it does not require a priori knowledge of the state nor a shared reference frame between the possibly remote observers and thus greatly simplifies the practical application .
we describe two schemes where the first one essentially can be seen as a direct implementation of schmidt decomposition , which identifies the maximal correlation directly . for bipartite
pure systems , this approach can conceptually be divided into two stages : ( i ) calibration that establishes the experimental schmidt decomposition @xcite of a pure state by local measurements and suitable filtering and ( ii ) two correlation measurements to verify the entanglement criterion . the second scheme shows how to use a decision tree to obtain a rapid violation of the threshold thereby identifying entanglement .
_ entanglement criterion._for a two - qubit quantum state @xmath0 , alice and bob observe correlations between their local pauli measurements @xmath1 and @xmath2 , respectively .
they are defined as the expectation values of the product of the two measurements , @xmath3 $ ] , with the so called correlation tensor elements @xmath4 $ ] . the local values @xmath5
( @xmath6 ) , with @xmath7 being the identity operator , form the local bloch vector of alice ( bob ) . using these measurements a sufficient condition for entanglement can be formulated as @xcite : @xmath8 for pure states this is also a necessary condition , while for mixed states care has to be taken .
for those the likelihood for detecting the entanglement decreases with purity .
an extension of ( [ simple_crit ] ) can generally identify entanglement of an arbitrary mixed state , however , then loosing the state independence @xcite .
note two important facts .
first , eq . can be seen as a state independent entanglement witness derived without any specific family of entangled states in mind .
second , to test whether the state is entangled , it is sufficient to break the threshold , i.e. , it is neither required to measure all correlations nor to compute the density matrix . )
generally scales as the one for full state tomography , we can directly evaluate the criterion from raw data without any further numerics required to reconstruct a physical density matrix @xcite ] rather , it is now the goal to find strategies which minimize the number of correlation measurements .
we show how this can be done by a particularly designed decision tree , or by identifying a schmidt decomposition from local results and filtering when necessary . _
schmidt decomposition._consider pure two qubit states . any such state has a schmidt decomposition @xmath9 , \label{schmidt}\ ] ] where the coefficients are real and the local bases @xmath10 and
@xmath11 are called the schmidt bases .
once the bases are known , alice constructs her local measurements @xmath12 and @xmath13 , and so does bob in analogy
. they can now detect entanglement with only _ _
t__wo correlation measurements because @xmath14 for all pure entangled states .
note , the laboratories are not required to share a common reference frame . in order to extract the schmidt bases from experimental data
one starts with local measurements , determining the local bloch vectors @xmath15 of alice ( bob ) ( those elements are related to the correlation tensor coefficients via @xmath16 ) .
we consider two cases .
first , suppose that a pure state has non - vanishing local bloch vectors .
their directions define the schmidt bases of alice and bob up to a global phase @xmath17 .
writing these bases in the computational basis @xmath18 we see that the required coefficients can be inferred directly from the local bloch vectors , @xmath19 on alice s side , and similarly for bob .
the global phase of @xmath20 shows up as the relative phase in the decomposition ( [ schmidt ] ) , i.e. @xmath21 ( with @xmath22 ) .
it can be determined , for example , from the @xmath23 correlation as @xmath24 . if bob would use the basis @xmath25 to build his observables @xmath26 and @xmath27 , the corresponding correlations @xmath28 would vanish for @xmath29 and the two measurements @xmath30 and @xmath31 would not suffice to detect entanglement . in such a case , however , the other two correlations , @xmath32 and @xmath33 , are non - zero , and can be used to reveal entanglement .
therefore , the determination of @xmath17 in the calibration is not essential if one accepts possibly one more correlation measurement .
second , in case of vanishing local bloch vectors , the pure state under consideration @xmath34 is maximally entangled and admits infinitely many schmidt decompositions . in order to truly prove entanglement bob
can thus freely choose some basis , say computational basis , for which the state will now be of the form @xmath35 .
the basis of alice can be found after _ filtering _ by bob in his schmidt basis : @xmath36 ( for an actual implementation see experimental section ) . when bob informs alice that his detector behind the filter clicked , the initial state becomes @xmath37 note that due to filtering a nonvanishing local bloch vector emerges for alice .
thus the respective schmidt basis can be found with the method described above and be used for the evaluation of @xmath38 .
directions randomly and independently , which effectively gives rise to a basis @xmath39 } for alice and @xmath40 for bob ( not detailed in the figure or the main text ) .
the scheme starts with measuring @xmath41 and follows at each step along the dashed arrow if the modulus of correlation is less than @xmath42 and along the continuous arrow otherwise .
the algorithm succeeds as soon as @xmath43 .
the measurements in blue shaded area suffice to detect all maximally entangled pure states with schmidt bases vectors along @xmath44,@xmath45 or @xmath46.,scaledwidth=28.0% ] _ decision tree._our second algorithm for entanglement detection does not even require any calibration and also applies directly to mixed states .
alice and bob choose three orthogonal local directions @xmath47 and @xmath46 independently from each other and agree to only measure correlations along these directions . in fig .
[ fig_dt ] we show exemplarily which correlations should be measured in order to detect entanglement in a small number of steps . starting with a measurement of @xmath41 , one continues along the solid ( dotted ) arrow , if the correlation is higher ( lower ) than some threshold value ( e.g. 1/2 in fig .
[ fig_dt ] ) .
the tree is based on the principle of correlation complementarity @xcite : in quantum mechanics there exist trade - offs for the knowledge of dichotomic observables with corresponding anti - commuting operators . for this reason ,
if the correlation @xmath48 is big , correlations @xmath49 and @xmath50 have to be small because their corresponding operators anti - commute with the operator @xmath51 .
therefore , the next significant correlations have to lie in the @xmath52 plane of the correlation tensor and thus the tree continues with a measurement of the @xmath23 correlation .
this concept can be generalized to multiqubit states , a decision tree for three qubits is given in the appendix .
the number of detected states grows with the number of steps through the decision tree .
since condition ( 1 ) is similar to the purity of a state , obviously , the scheme succeeds the faster the more pure a state is ( see appendix for detailed analysis ) .
varying the threshold value does not lead to any significant changes in the statistic of detected states . finally , we connect both methods discussed here for the analysis of multiqubit states .
a numerical simulation for pure states reveals that the correlation measurement along local bloch vectors gives correlations close to the maximal correlations in more than @xmath53 of the cases .
therefore , these local directions give an excellent starting point for the decision tree .
is created by type i spdc process .
an yttrium vanadate crystal ( yvo@xmath54 ) is used to manipulate the phase @xmath55 of the prepared state . for preparation and analysis of the state half- ( hwp ) and quarter ( qwp ) waveplates are employed .
brewster plates ( bp ) can be introduced to make the state asymmetric and to perform the filter operation , respectively.,scaledwidth=40.0% ] _
experiment._ for the demonstration of these new simple analysis methods we first use two photon polarization entangled states . in the following we will thus replace the computational basis states by horizontal ( @xmath56 ) and vertical ( @xmath57 ) linear polarization , respectively .
the photon source ( see fig .
[ source ] ) is based on the process of spontaneous parametric down conversion ( spdc ) using a pair of crossed type i cut @xmath58-barium - borate ( bbo ) crystals pumped by a cw laser diode at a wavelength of @xmath59 nm with linear polarization of @xmath60 .
it emits pairs of horizontally and vertically polarized photons which superpose to the state @xmath61 @xcite .
the spectral bandwidth of the photons is reduced to @xmath62 nm using interference filters and two spatial emission modes are selected by coupling the photon pairs into two separate single mode fibers . for the purpose of preparing any pure two - qubit state
, the polarization of each photon can be rotated individually by a set of quarter- ( qwp ) and half ( hwp ) waveplates in each mode . by tilting an yttrium vanadate crystal ( yvo@xmath54 ) in front of the bbos , the relative phase @xmath55 between the photon pairs can be set .
additionally , the state can be made asymmetric by removing a portion of vertically polarized light in one spatial mode with a brewster plate ( bp ) . in the last step of the experiment ,
the polarization of each photon is analyzed with additional half- and quarter waveplates and projection on @xmath63 and @xmath64 using a polarizing beam splitter ( pbs ) .
the local filtering of a maximally entangled state can be accomplished by placing a brewster plate in front of the analysis waveplates .
this brewster plate reflects with a certain probability vertically polarized photons and together with detection of a photon behind the brewster plate implements the filtering operation ( [ filtering ] ) .
finally , the photons are detected by fiber - coupled single photon detectors connected to a coincidence logic . _
experimental schmidt decomposition._ let us consider the state shown in figure [ schmidtsym]a ) .
the protocol starts with alice and bob locally measuring the polarization of the photons enabling them to individually determine the local bloch vectors . for high efficiencies , e.g. , possible in experiments with atoms or ions , the local measurements can indeed be done independently . and @xmath65 .
] if nonvanishing local bloch vectors can be identified , one can proceed to the next step .
for the example here , the local expectation values are close to zero and filtering has to be applied . by using a brewster plate in front of bob s analysis waveplate and under the condition that the filtering operation is successful , local bloch vectors emerge ( see fig .
[ schmidtsym]b ) , in this case we obtain @xmath66 and @xmath67 . in the next step alice and
bob use their local bloch vectors to realign their analyzers to the new local schmidt bases @xmath68 , @xmath69 and @xmath70 , @xmath71 , respectively .
this diagonalizes the correlation tensor as depicted in figure [ schmidtsym]c ) .
therefore , it is only necessary to measure @xmath72 and @xmath73 to prove entanglement since @xmath74 .
hence , @xmath75 _ local _ measurements are needed in the first step of the algorithm , if necessary three combined measurements are needed for filtering , and finally only two correlation measurements have to be performed for entanglement detection .
_ application of the decision tree._ in order to demonstrate the application of the decision tree we will apply it to three states . for the first state @xmath76 , whose correlation tensor is depicted in [ hhvvandrrll]a ) the decision tree ( see fig . [ fig_dt ] ) starts with the measurement of the correlation @xmath77 and continues with @xmath78 .
these two measurements already prove entanglement since @xmath79 . for a ) and @xmath80 for b).,scaledwidth=40.0% ] for a second state , @xmath81 , ( see fig .
[ hhvvandrrll]b ) , we obtain a correlation of @xmath82 , close to zero .
consequently , the next steps according to our algorithm ( fig . [ fig_dt ] ) are to determine the correlation @xmath83 followed by @xmath84 , with their squares adding up to a value of @xmath85 and hence proving entanglement . as a last example we consider the initial state of fig .
[ schmidtsym ] .
according to our decision tree we need to measure @xmath86 , @xmath87 and @xmath88 , thus giving a value of @xmath89 and proving entanglement with only three steps . _ many qubits._for the demonstration of multiqubit entanglement detection
, we use two three - photon polarization entangled states : the @xmath90 state @xcite and the @xmath91 state @xcite ( see fig .
[ multi ] ) . in order to observe these states , a collinear type ii spdc source together with a linear setup to prepare the four - photon dicke state @xmath92
is used @xcite . once the first photon is measured to be vertically polarized , the other three photons are projected into the @xmath90 state .
similarly , the three - photon @xmath91 state is obtained if the first photon is measured to be @xmath93 polarized . and @xmath90 state .
the corresponding fidelities are equal to 92.23% and 89.84%.,scaledwidth=40.0% ] the protocol for entanglement detection starts with observers locally measuring the polarization of the photons enabling them to individually determine the local bloch vectors . for the @xmath91 state we obtain @xmath94 , @xmath95 and @xmath96 .
the local bloch vectors suggest that the correlation @xmath97 is big .
therefore the decision tree starts with the measurement of @xmath98 and continues with @xmath99 ( see fig . 7 in the appendix ) .
these two measurements already prove entanglement because @xmath100 . for the @xmath90 state ,
the local bloch vectors : @xmath101 , @xmath102 and @xmath103 suggest that now the correlation @xmath104 is big . indeed , we observe @xmath105 .
the decision tree is the same as above but with local axes renamed as follows @xmath106 .
therefore , the second measurement has to be @xmath107 . with @xmath108
we again prove entanglement as @xmath109 . _
conclusions._ we discussed and experimentally implemented two methods for fast entanglement detection for states about which we have no a priori knowledge .
they are well suited for quantum communication schemes as the parties do not have to share a common reference frame , making the scheme insensitive to a rotation of the qubits during their transmission to the distant laboratories .
the two methods use a particularly simple and practical entanglement identifier @xcite .
one of them can be seen as experimental schmidt decomposition and the other establishes a sequence of correlation measurements leading to entanglement detection in a small number of steps .
we thank m. ukowski for stimulating discussions .
this work is supported by the eu project qessence , the daad / mnisw , the dfg - cluster of excellence map , and by the national research foundation and ministry of education in singapore .
wl is supported by the mnisw grant no .
n202 208538 and by the foundation for polish science .
cs thanks qccc of the elite network of bavaria for support . |
we are concerned here with the issue of classifying documents into categories .
more precisely , we begin with a number of categories ( e.g. , ` tennis , soccer , skiing ' ) , each already containing certain documents . our goal is to determine into which categories newly given documents ought to be assigned , and to do so on the basis of the distribution of each document s words .
many methods have been proposed to address this issue , and a number of them have proved to be quite effective ( e.g.,@xcite ) .
the simple method of conducting hypothesis testing over word - based distributions in categories ( defined in section 2 ) is not efficient in storage and suffers from the _ data sparseness problem _ , i.e. , the number of parameters in the distributions is large and the data size is not sufficiently large for accurately estimating them . in order to address this difficulty ,
@xcite have proposed using distributions based on what we refer to as _ hard clustering _ of words , i.e. , in which a word is assigned to a single cluster and words in the same cluster are treated uniformly .
the use of hard clustering might , however , degrade classification results , since the distributions it employs are not always precise enough for representing the differences between categories .
we propose here to employ _ soft clustering _ ,
i.e. , a word can be assigned to several different clusters and each cluster is characterized by a specific word probability distribution .
we define for each category a _ finite mixture model _ , which is a linear combination of the word probability distributions of the clusters .
we thereby treat the problem of classifying documents as that of conducting statistical hypothesis testing over finite mixture models . in order to accomplish hypothesis testing
, we employ the em algorithm to efficiently and approximately calculate from training data the maximum likelihood estimates of parameters in a finite mixture model .
our method overcomes the major drawbacks of the method using word - based distributions and the method based on hard clustering , while retaining their merits ; it in fact includes those two methods as special cases .
experimental results indicate that our method outperforms them .
although the finite mixture model has already been used elsewhere in natural language processing ( e.g. @xcite ) , this is the first work , to the best of knowledge , that uses it in the context of document classification .
a simple approach to document classification is to view this problem as that of conducting hypothesis testing over word - based distributions . in this paper
, we refer to this approach as the _ word - based method _
( hereafter , referred to as wbm ) . letting @xmath0 denote a vocabulary ( a set of words ) , and @xmath1 denote a random variable representing any word in it , for each category @xmath2 @xmath3
, we define its _ word - based distribution _ @xmath4 as a histogram type of distribution over @xmath0 .
( the number of free parameters of such a distribution is thus @xmath5 ) .
wbm then views a document as a sequence of words : @xmath6 and assumes that each word is generated independently according to a probability distribution of a category .
it then calculates the probability of a document with respect to a category as @xmath7 and classifies the document into that category for which the calculated probability is the largest .
we should note here that a document s probability with respect to each category is equivalent to the _ likelihood _ of each category with respect to the document , and to classify the document into the category for which it has the largest probability is equivalent to classifying it into the category having the largest likelihood with respect to it .
hereafter , we will use only the term likelihood and denote it as @xmath8 .
notice that in practice the parameters in a distribution must be estimated from training data . in the case of wbm ,
the number of parameters is large ; the training data size , however , is usually not sufficiently large for accurately estimating them .
this is the _ data sparseness problem _ that so often stands in the way of reliable statistical language processing ( e.g.@xcite ) .
moreover , the number of parameters in word - based distributions is too large to be efficiently stored . in order to address the above difficulty , guthrie et.al .
have proposed a method based on hard clustering of words @xcite ( hereafter we will refer to this method as hcm ) .
let @xmath9 be categories .
hcm first conducts hard clustering of words .
specifically , it ( a ) defines a vocabulary as a set of words @xmath0 and defines as clusters its subsets @xmath10 satisfying @xmath11 and @xmath12 @xmath13 ( i.e. , each word is assigned only to a single cluster ) ; and ( b ) treats uniformly all the words assigned to the same cluster .
hcm then defines for each category @xmath2 a distribution of the clusters @xmath14 @xmath15 .
it replaces each word @xmath16 in the document with the cluster @xmath17 to which it belongs @xmath18 .
it assumes that a cluster @xmath17 is distributed according to @xmath14 and calculates the likelihood of each category @xmath2 with respect to the document by @xmath19 .frequencies of words [ cols="<,^,^,^,^,^,^",options="header " , ] we considered the following questions : \(1 ) the training data used in the experimentation may be considered sparse .
will a word - clustering - based method ( fmm ) outperform a word - based method ( wbm ) here ?
\(2 ) is it better to conduct soft clustering ( fmm ) than to do hard clustering ( hcm ) ?
\(3 ) with our current method of creating clusters , as the threshold @xmath20 approaches 0 , fmm behaves much like wbm and it does not enjoy the effects of clustering at all ( the number of parameters is as large as in wbm ) .
this is because in this case ( a ) a word will be assigned into all of the clusters , ( b ) the distribution of words in each cluster will approach that in the corresponding category in wbm , and ( c ) the likelihood value for each category will approach that in wbm ( recall case ( 2 ) in section 3 ) .
since creating clusters in an optimal way is difficult , when clustering does not improve performance we can at least make fmm perform as well as wbm by choosing @xmath21 .
the question now is `` does fmm perform better than wbm when @xmath20 is @xmath22 ? '' in looking into these issues
, we found the following : \(1 ) when @xmath23 , i.e. , when we conduct clustering , fmm does not perform better than wbm for the first data set , but it performs better than wbm for the second data set . evaluating classification results on the basis of each individual category , we have found that for three of the nine categories in the first data set , fmm0.5 performs best , and that in two of the ten categories in the second data set fmm0.5 performs best .
these results indicate that clustering sometimes does improve classification results _ when we use our current way of creating clusters_. ( fig .
[ fig : corn ] shows the best result for each method for the category ` corn ' in the first data set and fig .
[ fig : grain ] that for ` grain ' in the second data set . )
\(2 ) when @xmath23 , i.e. , when we conduct clustering , the best of fmm almost always outperforms that of hcm .
\(3 ) when @xmath24 , fmm performs better than wbm for the first data set , and that it performs as well as wbm for the second data set . in summary , fmm always outperforms hcm ; in some cases it performs better than wbm ; and in general it performs at least as well as wbm .
8.2 cm 8.2 cm for both data sets , the best fmm results are superior to those of cos throughout .
this indicates that the probabilistic approach is more suitable than the cosine approach for document classification based on word distributions .
although we have not completed our experiments on the entire reuters data set , we found that the results with fmm on the second data set are almost as good as those obtained by the other approaches reported in @xcite .
( the results are not directly comparable , because ( a ) the results in @xcite were obtained from an older version of the reuters data ; and ( b ) they used stop words , but we did not . ) we have also conducted experiments on the susanne corpus data and confirmed the effectiveness of our method .
we omit an explanation of this work here due to space limitations .
let us conclude this paper with the following remarks : 1 .
the primary contribution of this research is that we have proposed the use of the finite mixture model in document classification .
experimental results indicate that our method of using the finite mixture model outperforms the method based on hard clustering of words .
experimental results also indicate that in some cases our method outperforms the word - based method _ when we use our current method of creating clusters_. our future work is to include : 1 . comparing the various methods over the entire reuters corpus and over other data bases , 2 . developing better ways of creating clusters . our proposed method is not limited to document classification ; it can also be applied to other natural language processing tasks , like word sense disambiguation , in which we can view the context surrounding a ambiguous target word as a document and the word - senses to be resolved as categories .
we are grateful to tomoyuki fujita of nec for his constant encouragement .
we also thank naoki abe of nec for his important suggestions , and mark petersen of meiji univ . for his help with the english of this text .
we would like to express special appreciation to the six acl anonymous reviewers who have provided many valuable comments and criticisms . | we propose a new method of classifying documents into categories . we define for each category a _ finite mixture model _ based on _ soft clustering _ of words .
we treat the problem of classifying documents as that of conducting statistical hypothesis testing over finite mixture models , and employ the em algorithm to efficiently estimate parameters in a finite mixture model .
experimental results indicate that our method outperforms existing methods . |
the concept of `` entropy '' ( from the greek @xmath12 , transformation ) , was introduced in @xmath13 by rudolf julius emmanuel clausius in the context of thermodynamics@xcite .
this was motivated by his studies on reversible and irreversible transformations , as a measure of the amount of energy in a physical system , that can not be used to perform work . more specifically , clausius defined _ change in entropy _ of a thermodynamic system , during some reversible process where a certain amount of heat @xmath14 is transported at constant temperature @xmath15 , as @xmath16
. we can consider _ entropy _ as the cornerstone of thermodynamics , since all the thermodynamical principles involve , directly or indirectly , this fundamental concept . the first connection between the macroscopic clausius entropy of a system and its microscopic configurations was done by ludwig boltzmann in @xmath17 @xcite .
studying the approach to equilibrium of an ideal " gas @xcite , he realized that the entropy could be related to the number of possible microstates compatible with the thermodynamic properties of the gas .
for an isolated system in its terminal stationary state ( _ thermal equilibrium _ ) , boltzmann observation can be expressed as @xmath18 where @xmath2 is a positive constant and @xmath19 the number of microstates consistent with the macroscopic state .
this famous equation , known as _ boltzmann principle _ , can be regarded as a pillar in the foundations of _ statistical mechanics _ , which aims to describe the thermodynamic state of a system through the statistical approach of its microstates . for systems which are not isolated but instead are in contact with some reservoir ( of heat , particles , etc . ) , it is possible to derive ( under some assumptions ) , from eq .
( [ s_b ] ) @xmath20 @xcite , the celebrated boltzmann - gibbs entropy @xmath21 where @xmath22 , subject to the normalization condition @xmath23 , is the probability of the microscopic configuration @xmath24 .
in particular , for equiprobability , @xmath25 , @xmath26 , then boltzmann - gibbs entropy ( [ s_bg ] ) reduces to ( [ s_b ] ) .
since it refers to microscopic states , the boltzmann principle should be derivable from microscopic dynamics .
however , the implementation of such calculation has not been yet achieved .
consequently , @xmath0 statistical mechanics still remains based on hypothesis such as boltzmann s _ stosszahlansatz _ ( molecular chaos hypothesis)@xcite and _ ergodicity _ @xcite .
it can be easily shown that entropy ( [ s_bg ] ) is _ nonnegative _ , _ concave _ , _ experimentally robust _ ( or _ lesche - stable_)@xcite , and leads to a _ finite entropy production per unit time_@xcite .
moreover , it is _ additive_. in other words , if @xmath27 and @xmath28 are two _ probabilistically independent _
subsystems , i.e. , @xmath29 , then it is straightforwardly verified that @xmath30 hence , if we have @xmath31 subsystems , @xmath32 , where the notation is self - explanatory .
more generally , when correlations are weak " enough , @xmath8 is _ extensive _ , i.e. , such that the @xmath33 is _ finite_. despite the lack of first - principle derivations , boltzmann - gibbs statistical mechanics has a history plenty of successes in the treatment of systems where short spatio / temporal interactions dominate .
this kind of interactions favor ergodicity and independence properties , necessary in khinchin s approach of @xmath8 @xcite .
consequently , it is perfectly plausible that physical entropies other than the boltzmann - gibbs one can be defined in order to suitably treat anomalous systems , for which ergodicity and/or independence are not verified . as examples of anomalies we can mention : metastable states in long - range interacting hamiltonian systems , metaequilibrium states in small systems ( i.e. , systems whose number of particles is much smaller than avogrado s number ) , glassy systems , some types of dissipative systems , systems that in some way violate ergodicity , and , _ last but not least _ , systems with non - markovian memory , like it seems to be the case of financial ones . generically speaking , systems that may possibly have a multifractal , scale - free or hierarchical structure in the occupancy of their phase space . motivated by this scenario , one of us proposed in @xmath34 the entropy@xcite @xmath35 which generalizes @xmath8 , such that @xmath36 , as the basis of a possible generalization of boltzmann - gibbs statistical mechanics@xcite and where the _ entropic index _
@xmath6 should be determined _ a priori _ from microscopic dynamics . just like @xmath8 ,
@xmath37 is _ nonnegative _ , _ concave _ , _ experimentally robust _ ( or _ lesche - stable _ ) ( @xmath38 ) , and leads to a _ finite entropy production per unit time _ ( @xcite and references therein ) .
moreover , it has been recently shown@xcite that it is also _ additive _ , i.e. , @xmath39 for special kinds of _ correlated _ systems , more precisley when phase - space is occupied in a scale - invariant form . since
its proposal , entropy ( [ sq ] ) has been the source of numerous results in both fundamental and applied physics as well as in other scientific areas such as biology , chemistry , economics , geophysics and medicine @xcite .
it is our purpose here to review some new results concerning applications to economics and more specifically to the theory of finances .
before discussing some specific quantities of financial interest , let us see the form of the probability density function naturally derived from the variational principle related to entropy ( [ sq ] ) .
we consider its continuous version , i.e. , @xmath40 ^{q}\ dx}{1-q}. \label{sq - cont}\ ] ] natural constraints in the maximization of ( [ sq - cont ] ) are @xmath41 corresponding to normalization , and @xmath42 ^{q}}{\int \left [ p\left ( x\right ) \right ] ^{q}dx}\ dx\equiv \left\langle x\right\rangle _ { q}=\bar{\mu } _ { q } \ , , \ ] ] @xmath43 ^{q}}{\int \left [ p\left ( x\right ) \right ] ^{q}dx}\ dx\equiv \left\langle \left ( x-\bar{\mu}_{q}\right ) ^{2}\right\rangle _ { q}=\bar{\sigma } _
{ q}^{2}\,,\ ] ] corresponding to the _ generalized _ mean and variance of a relevant quantity @xmath44 , respectively .
it is noteworthy that , averages weighted with a function of the probabilities as in ( [ mean])-([variance ] ) allow to mimic the way individuals behave in face to risky choices@xcite ( see also @xcite ) .
in fact , the _ prospect _ theory proposed by kahneman and tversky@xcite for analyzing decision - making under risk , is founded on the concept of `` decision weights '' that can be modeled by akin functional forms@xcite . from the variational problem for ( [ sq - cont ] ) under the above constraints , one obtains @xmath45 ^{\frac{1}{1-q}},\qquad \left ( q<3\right),\ ] ] where , @xmath46 } { \gamma \left [ \frac{2-q}{1-q } \right ] } \sqrt{\frac{1-q}{\pi } \mathcal{b}_{q } } & \leftarrow & q<1 \\[3 mm ] \frac{\gamma \left [ \frac{1}{q-1}\right ] } { \gamma \left [ \frac{3-q}{2q-2 } \right ] } \sqrt{\frac{q-1}{\pi } \mathcal{b}_{q } } & \leftarrow & q>1 \end{array } % \right . , \ ] ] and @xmath47 ^{-1}.\ ] ] the upper bound @xmath48 guarantees normalizability . defining the @xmath6-_exponential _ function as @xmath49 ^ { \frac{1}{1-q}}\qquad \left ( e_{1}^x \equiv e^{x}\right ) , \ ] ] ( @xmath50 if @xmath51 ) we can rewrite the probability density ( [ pq-1 ] ) as @xmath52 hereon referred to as @xmath6-_gaussian _ probability density function . for @xmath53
, the @xmath6-gaussian form recovers the student s @xmath54-distribution with @xmath55 degrees of freedom ( @xmath56 ) .
consistently , for @xmath57 ( [ pq ] ) presents an asymptotic _ power - law _ behavior . also , if @xmath58 with @xmath59 , @xmath60 recovers the @xmath61-distribution with @xmath62 degrees of freedom .
consistently , for @xmath63 , @xmath64 has a _ compact support _ which is defined by the condition @xmath65 .
many other entropic forms have been introduced in the literature for various interesting purposes .
one of the advantages of entropy ( [ sq ] ) is that it yields power - law tails , which play a particularly relevant role , as well known .
let us recall succinctly the two basic central limit theorems : ( 1 ) a convoluted distribution with a finite second moment approaches , in the limit of @xmath66 convolutions , a gaussian attractor ; ( 2 ) a convoluted distribution with a divergent second moment , approaches , in the same limit , a lvy distribution @xmath67 ( with @xmath68)@xcite .
however , through dynamics different from the convolution one , for instance with some sort of memory across successive steps ( i.e. , _ nonindependence _ of the successive distributions ) , different ubiquitous distributions might arise ( see also @xcite ) .
if the laplacian term in the linear diffusion equation is a standard second derivative or a fractional derivative , the gaussian or the lvy distributions are respectively attained as solutions .
however , much more complex and rich dynamics clearly exist in nature , for example those associated with a variety of nonlinear fokker - planck equations involving nontrivial correlations , multiplicative noise or other effects , as we will see in the following section .
moreover , simple convolutions allow only for asymptotic behavior like the @xmath11 ( gaussian ) or @xmath69 ( lvy distributions ) ones .
but they do _ not _ allow fat - tailed distributions associated with @xmath70 .
however , many complex systems in nature as well as in social sciences exhibit exponents which precisely belong to that interval . an example of financial interest is exhibited in fig [ fig - dj1 ] .
, _ versus _ return , @xmath61 .
the symbols represent @xmath71 for the dow jones industrial daily return index from 1900 until 2003 .
the solid line represents the best q - gaussian numerical adjust with @xmath72 and @xmath73 ( as obtained in @xcite ) and the dashed line a gaussian fit . ]
the gaussian distribution , recovered in the limit @xmath74 of expression ( [ pq ] ) , can be derived on a variety of grounds . for instance , it has been derived , through arguments based on the dynamics , by l. bachelier in his 1900 work on price changes in paris stock market , and also by a. einstein in his 1905 article on brownian motion . in particular , starting from a langevin dynamics , one is able to write the corresponding fokker - planck equation and , from it , to obtain as solution the gaussian distribution .
analogously , it is also possible , from certain classes of stochastic differential equations and their associated fokker - planck equations , to obtain the distribution given by ( [ pq ] ) . in this section
we will discuss dynamical mechanisms which lead to probability functions with asymptotic power - law behavior of the @xmath6-gaussian form .
microscopic dynamics containing multiplicative noise may be encountered in many dynamical processes and , due to its significance , has been the subject of numerous studies in the last decades .
the presence of additive noise , besides the multiplicative contribution , is in fact a quite realistic ingredient since not all the fluctuations are processed multiplicatively . in previous work@xcite , we considered a class of differential stochastic equations of the form @xmath75 where @xmath76 are arbitrary functions of the stochastic variable @xmath44 , and @xmath77 , two independent zero - mean gaussian white noises with variance @xmath78 and @xmath79 , respectively .
we have shown that for special forms of the deterministic force , namely @xmath80 , the stationary probability density functions ( by using the it prescription ) are of the form @xmath81
^ 2}\ , , \ ] ] where @xmath82 and @xmath83 from the point of view of entropy @xmath7 , the density function ( [ ps ] ) derives from the variational principle , under the constraints of normalization and generalized variance of @xmath84 .
in particular , @xmath6-gaussian distributions can be derived from a stochastic process of the linear form@xcite @xmath85 in fact , its associated fokker - planck equation is @xmath86}{\partial x^2}\ ] ] that has the alternative form @xmath87 where @xmath88 , and @xmath89 is a constant related to the other model parameters .
( [ nlfp ] ) is a _ nonlinear _ diffusion equation , familiarly known as porous media equation .
its steady state solution has the form @xmath90 with @xmath6 and @xmath91 defined as above . in the particular case
@xmath92 the standard gaussian steady state is obtained , corresponding to a purely additive process .
taking into account that empirical returns where found to follow a @xmath6-gaussian distribution@xcite ( see also fig . [ fig - dj1 ] ) , eq .
( [ langevin_gen ] ) , complemented by the it prescription , provides a simple mechanism to model the dynamics of prices . along similar lines
it has been worked out , for instance , an option - pricing model which is more realistic than the celebrated black - scholes one@xcite ( recovered as the @xmath11 particular case ) .
the @xmath6-exponential character of the solutions of eq .
( [ nlfp ] ) is not exclusive of the steady state but it also emerges along the time evolution@xcite . in the presence of multiplicative noise , the system variables directly couple to noise .
therefore , behaviors are observed that can not occur in the presence of additive noise alone . on the other hand
the additive noise plays a fundamental role allowing the existence of a reasonable and normalizable steady state by avoiding collapse of the distribution at the origin . the particular interplay between additive and multiplicative noises , as well as that between deterministic and stochastic drifts
, can lead to the appearance of @xmath6-exponential forms .
the @xmath6-exponential distributions include the boltzmann - gibbs one as a special case ( @xmath11 ) .
while the latter corresponds to the standard thermal equilibrium , the @xmath93 is expected to be related to quasi - stationary or long - living out - of - equilibrium regimes .
intricate dynamical behavior is a common feature of many non - equilibrium systems , which can be also characterized by power - law probability density functions . to this class
belong systems whose dynamical behavior shows spatio / temporal fluctuations of some intensive quantity
. this quantity may be the inverse temperature , like in the case of interactions of hadrons from cosmic rays in emulsion chambers ; the energy dissipation in turbulent fluids , or simply the width of some white noise , as assumed in many financial models , such as in the famous heston model@xcite .
the connection between this sort of dynamics and nonextensive entropy was first made by g. wilk and z. wodarczyk @xcite and later extended by c. beck and e.g.d .
cohen@xcite , who called it _
superstatistics_. in this statistics of statistics " , beck and cohen aimed to treat non - equilibrium systems from the point of view of long - living stationary states characterized by a temporally or spatially fluctuating intensive parameter . such condition can be mathematically expressed by @xmath94=\int_{0}^{\infty } f\left ( \beta \right ) \,e^{-\beta e(z)}\,d\beta \label{superstatistics}\ ] ] where @xmath95 $ ] is a kind of effective boltzmann factor , @xmath96 a function of some relevant variable @xmath97 , and @xmath98 the probability density function of the inverse temperature @xmath91 .
superstatistics is intimately connected with nonextensive statistical mechanics .
more precisely , it is possible to derive a generalized boltzmann factor which is exactly @xmath99 $ ] , when @xmath98 is the gamma distribution , i.e. , @xmath100 } \left ( \frac{\beta } { c}\right ) ^{c-1}e^{-\beta /b}\ , , \ ] ] where the @xmath6-exponential functional form of the effective boltzmann factor turns clearly visible its asymptotic power - law behavior .
it is noteworthy that the above effective boltzmann factor is also a good approximation for other @xmath98 probability density functions @xcite .
the fluctuating character of volatility in financial markets has been considered , since a few decades ago , as major responsible for price change dynamics @xcite .
in fact , the intermittent character of return time series is usually associated with localized bursts in volatility and thus called _ volatility clustering @xcite_. the temporal evolution of the second - order moment , known as _
heteroskedasticity @xcite _ , has proven to be of extreme importance in order to define better performing option - price models @xcite the first proposal aiming to modelize and analyze economical time series with time - varying volatility was made by r.f .
engle @xcite , who defined the _ autoregressive conditional heteroskedasticity _ ( @xmath103 ) process . in his seminal article
, engle stated that a heteroskedastic observable @xmath97 ( e.g. , the _ return _ ) would be defined as @xmath104 where @xmath105 represents an independent and identically distributed stochastic process with null mean and unitary variance ( @xmath106 , @xmath107 ) associated to a probability density function @xmath108 , and @xmath109 the volatility . in the same work , engle also suggested a simple dynamics for volatilities , a linear dependence of @xmath110 on the @xmath62 previous values of @xmath111 $ ] , @xmath112 afterwards named as @xmath113 _ linear process @xcite_. the @xmath113 process is uncorrelated and for @xmath114 it presents a volatility autocorrelation function with a characteristic time of order @xmath115 @xcite . in order to give a more flexible structure to the functional form of @xmath116 , t. bollerslev generalized eq .
( [ arch - def ] ) defining the @xmath117 process@xcite @xmath118 which reduces to @xmath113 process , when @xmath119 . for the @xmath120 ( @xmath121 and @xmath122 )
, we can straightforwardly determine the @xmath123 moment of the _ stationary _ probability density function @xmath124 , particularly its second moment @xmath125 and the fourth moment , which equals the kurtosis @xmath126 , @xmath127 for processes with unitary variance , i.e. , @xmath128 .
continuous approaches are becoming more often used ( mainly in the treatment of high - frequency data : see , e.g. , @xcite ) .
moreover , @xmath103-like processes fail in reproducing the volatility autocorrelation function power law behavior@xcite .
but , despite these facts , the @xmath103 family of processes ( particularly @xmath129 and @xmath130 ) is still considered a cornerstone in econometrics due to its simplicity and satisfactory efficiency in financial time series mimicry . having a glance at eq .
( [ hetero - def ] ) , we can verify that the distribution @xmath131 of the stochastic variable @xmath97 has , at each time step @xmath54 , the same functional form of the noise distribution , @xmath132 , but with a variance @xmath109 .
this property allows one to look at process @xmath133 as a process similar to those occuring in some non - equilibrium systems with a longstanding stationary state .
specifically , this principle has allowed to establish , firstly for @xmath134 @xcite and then for @xmath135 @xcite , a connection between @xmath136 and @xmath137 , @xmath138 and , @xmath108 , the latter assumed to be of the following @xmath139-gaussian form @xmath140 ^{\frac{1}{q_{n}-1 } } } = a_{q_n}\,e_{q_n}^{-\omega^2/(5 - 3q_n ) } , \qquad \left ( q_{n}<\frac{5}{3}% \right ) , \label{pqn}\ ] ] ( upper plots ) and @xmath141 ( lower plots ) with @xmath128 . for each plot
the symbols represent the relative frequency , @xmath142 , and the line the corresponding probability function , @xmath143 . in ( a ) @xmath144 @xmath145 ; ( b ) @xmath146 @xmath147 ; ( c ) @xmath148 @xmath149 ; ( d ) @xmath150 @xmath151.,title="fig : " ] ( upper plots ) and @xmath141 ( lower plots ) with @xmath128 . for each plot
the symbols represent the relative frequency , @xmath142 , and the line the corresponding probability function , @xmath143 . in ( a ) @xmath144 @xmath145 ; ( b ) @xmath146 @xmath147 ; ( c ) @xmath148 @xmath149 ; ( d ) @xmath150 @xmath151.,title="fig : " ] ( upper plots ) and @xmath141 ( lower plots ) with @xmath128 . for each plot
the symbols represent the relative frequency , @xmath142 , and the line the corresponding probability function , @xmath143 . in ( a ) @xmath144 @xmath145 ; ( b ) @xmath146 @xmath147 ; ( c ) @xmath148 @xmath149 ; ( d ) @xmath150 @xmath151.,title="fig : " ] ( upper plots ) and @xmath141 ( lower plots ) with @xmath128 . for each plot
the symbols represent the relative frequency , @xmath142 , and the line the corresponding probability function , @xmath143 . in ( a ) @xmath144 @xmath145 ; ( b ) @xmath146 @xmath147 ; ( c ) @xmath148 @xmath149 ; ( d ) @xmath150 @xmath151.,title="fig : " ] by making the ansatz @xmath152 @xmath153 , where @xmath154 is the @xmath6-gaussian probability density function which maximizes @xmath37 ( eq .
( [ sq ] ) ) , and by imposing the matching of second ( @xmath128 ) and fourth order moments , it is possible to establish , for @xmath155 , a relation containing the dynamical parameters @xmath136 and @xmath137 and entropic indexes @xmath6 and @xmath139 : @xmath156}-c(q - q_{n}).\ ] ] for @xmath157 , eq .
( [ garch - rel ] ) reduces to the one corresponding to @xmath129 , @xmath158 and , for @xmath159 , one has @xmath160 . the validity of eqs .
( [ garch - rel ] ) and ( [ arch - rel ] ) is depicted in fig .
[ fig - arch ] .
the discrepancy between @xmath154 and @xmath152 can be evaluated by computing the sixth - order moment percentual difference , which is never greater than @xmath161 @xcite . since @xmath162 and @xmath163 , for @xmath164 we can write @xmath165 as the conditional probability density function of @xmath97 given @xmath166 . considering that , @xmath167 and @xmath152 @xmath153
, we obtain the stationary probability density function for squared volatility@xcite , @xmath168 } \;e^ { - \frac{1}{2 \kappa \sigma^{2 } } } , \ ] ] where @xmath169 as one can observe in fig .
[ fig - vol ] , the ansatz gives a quite satisfactory description for @xmath166 probability density function , suggesting a connection between the @xmath103 class of processes and nonextensive entropy .
these explicit formulas can be helpful in applications related , among others , to option prices , where volatility forecasting plays a particularly important role .
, numerically obtained for a gaussian noise with @xmath170 and the gray line the respective inverse cumulative distribution , @xmath171 with @xmath172 for @xmath173 . ]
albeit uncorrelated , stochastic variables @xmath174 are _ not _ independent .
applying the @xmath6-generalized kullback - leibler relative entropy @xcite to stationary joint probability density function @xmath175 and @xmath176 @xmath177 ^{2}$ ] it is possible to quantify the degree of dependence between successive returns , through an optimal entropic index , @xmath178 . in ref .
@xcite , it was verified the existence of a direct relation between dependence , @xmath179 , the non - gaussianity , @xmath6 , and the nature of the noise , @xmath139 .
an interesting property emerged , namely that , whatever be the pair @xmath180 that results in a certain @xmath6 for the stationary probability density function , one obtains the same @xmath179 and , consequently , the time series will present the same degree of dependence @xcite . see fig .
[ fig - qopq ] .
@xmath6 for typical @xmath181triplets .
the arrow points two examples which were obtained from _
different _ triplets and nevertheless coincide in what concerns the resulting point @xmath182 . ]
it was also verified ( see fig .
[ fig - quantf ] ) for @xmath129 that the degree of dependence varies visibly with @xmath136 and with the lag @xmath183 between returns .
this dependence would be related to a short - memory in volatility @xcite .
the variance between these results and the empirical evidence of persistence in the real return time series dependence degree for time intervals up to @xmath184 days recently found , shows that financial markets dynamics are , in fact , ruled by some form of long - memory processes in volatility @xcite . _
@xmath136 for @xmath129 process . even for large allowed @xmath136 values ,
the decrease of dependence degree ( i.e. , increase of @xmath179 ) is visible when the time lag increases , which is not compatible with the empirical evidence of dependence degree around @xmath185 verified in dow jones return time series for time lags up to 100 days , panel ( b ) .
comparing the two figures ( a ) and ( b ) , it appears that financial markets present long - memory in the volatility similar to a @xmath186 process with @xmath187.,title="fig : " ] _ versus _
@xmath136 for @xmath129 process . even for large allowed @xmath136 values ,
the decrease of dependence degree ( i.e. , increase of @xmath179 ) is visible when the time lag increases , which is not compatible with the empirical evidence of dependence degree around @xmath185 verified in dow jones return time series for time lags up to 100 days , panel ( b ) .
comparing the two figures ( a ) and ( b ) , it appears that financial markets present long - memory in the volatility similar to a @xmath186 process with @xmath187.,title="fig : " ] let us comment at this point that the connection between several entropic indexes , each one related to a different observable , is compatible with the dynamical scenario within which the nonextensive statistical mechanics is formulated .
in fact , several entropic indexes emerge therein , coalescing onto the same value @xmath11 when ergodicity is verified .
another important observable in financial markets is the traded volume @xmath188 ( the number of shares of a stock traded in a period of time @xmath15 ) . in ref .
@xcite it was proposed an ansatz for the probability density function of high - frequency traded volume , which presents two power - law regimes ( small and large values of @xmath188 ) , @xmath189 where @xmath190 represents the traded volume expressed in its mean value unit @xmath191 , i.e. , @xmath192 , @xmath193 and @xmath194 are positive parameters and @xmath195 .
the probability density function ( [ prob - obt ] ) was recently obtained from a mesoscopic dynamical scenario @xcite based in the feller process @xcite ( using it definition ) , @xmath196 where instead of being constant in time , @xmath197 varies on a time scale much larger than the time scale of order @xmath198 needed to reach stationarity .
the deterministic term of eq .
( [ eq - vol1 ] ) represents a restoring market mechanism which tries to keep the traded volume in some typical value @xmath199 and the second term reflects stochastic memory and , basically , the effect of large traded volumes .
in fact , large values of @xmath190 will provoke large amplitude of the stochastic term , leading to an increase or decrease of the traded volume ( stirred or hush stock ) depending on the sign of @xmath200 .
the fluctuation of @xmath91 , alike to fluctuations in the mean value of @xmath190 , can be related with changes in the activity volume due to agents herding mechanism caused by price movements or news . solving the corresponding fokker - planck equation for eq .
( [ eq - vol1 ] ) we got the conditional probability of @xmath190 given @xmath91 , @xmath201 } \left ( \beta \,v\right ) ^{\alpha } \,\exp \left ( -\beta \,v\right ) \qquad ( \alpha > -1,\ \beta > 0 ) .
\label{pvb}\ ] ] since @xmath202 considering that @xmath197 follows a gamma distribution , @xmath203 } % \left ( \frac{\beta } { \lambda } \right ) ^{\delta -1}\exp \left ( -\frac{\beta } { % \lambda } \right ) \qquad ( \delta > 0,\ \lambda > 0 ) , \label{pb}\ ] ] eq .
( [ super ] ) yields , @xmath204 where , @xmath205 and @xmath206 .
a numerical simulation for @xmath207 minute traded volume in nyse stock index is exhibited in fig .
[ fig - ny ] . ) .
parameters are @xmath208 , @xmath209 , @xmath210 and @xmath211.,title="fig : " ] ) .
parameters are @xmath208 , @xmath209 , @xmath210 and @xmath211.,title="fig : " ] another possible mechanism to describe the dynamics of volumes , has been recently proposed@xcite on the basis of mean - reverting processes with additive - multiplicative structure , namely , @xmath212 where , @xmath193 , @xmath213 are positive constants and @xmath214 , @xmath215 independent standard wienner processes . that is , following the ideas presented in section [ sec : additive - multiplicative ] , in addition to fluctuations endogenously processed by the market ,
other fluctuations are taken into account that affect the dynamics directly .
the stationary solution of the fokker - planck equation associated to the it - langevin eq .
( [ eq - vol2 ] ) has the form ( [ pvfinal ] ) .
moreover , the underlying sequences present intermittent bursts ( similar to those exhibited in fig .
[ fig - ny].b ) and , preliminary results indicate the presence of persistent power - law correlations , as observed in real data sequences .
( [ eq - vol2 ] ) belongs to a larger class of processes with two - fold power - law behavior that can be also suitable for volumes , as well as , for other mean - reverting variables such as volatilities .
additive - multiplicative processes are at the core of nonextensive statistical mechanics . in the same natural way that standard
brownian motion leads to gaussianity , linear additive - multiplicative stochastic processes lead to @xmath6-gaussian distributions .
special ( scale - invariant ) correlations , that forbid convergence to the usual gauss or lvy limits , lead to a new type of statistical distributions .
a remarkable feature is that the resulting power - law distributions may have exponents out of the lvy range , thus allowing to embrace a larger variety of empirical processes .
the presence of two gaussian white noises , one either enhanced or reduced by internal information , and another purely exogenous , represents quite realistic features present in a variety of systems , thus justifying the ubiquity of the probability distributions associated to such kind of processes .
in particular , as we have shown , they allow to model the dynamics of prizes , volatilities , stock - volumes and other relevant financial observables .
another expression of the mechanism underlying the nonextensive theory is connected to the existence of a fluctuating intensive parameter ( or `` inverse temperature '' ) following the ideas that foundate the beck and cohen superstatistics @xcite .
we have shown that these principles allow an alternative description of the dynamics of stock - volumes .
furthermore , such kind of mechanism allows an interesting perspective for treating the family of _ g / arch _ processes .
the fact that the resulting probability density functions can be described in terms of @xmath6-gaussian distributions , provides a tractable way of dealing with empirical distributions that match the _
g / arch _ types .
some of the discussions presented in this review have been done at a mesoscopic scale .
the determination , from more microscopic models , of the parameters used at the mesoscopic scale is certainly welcome .
curado e.m.f . and
tsallis c. , j. phys . a * 24 * , l69 ( 1991 ) ; corrigenda : 24 , 3187 ( 1991 ) and 25 , 1019 ( 1992 ) ; a 24 , l69 ( 1991 ) ; tsallis c. , mendes r. s. and plastino a.r . , physica a * 261 * , 534 ( 1998 ) ; see http://tsallis.cat.cbpf.br/biblio.htm for an updated bibliography . c. tsallis , proceedings of the 31st workshop of the international school of solid state physics
complexity , metastability and nonextensivity " , held at the ettore majorana foundation and centre for scientific culture ( erice , july 2004 ) , eds . c. beck , a. rapisarda and c. tsallis ( world scientific , singapore , 2005 ) , in press [ cond - mat/0409631 ] ; y. sato and c. tsallis , proceedings of the summer school and conference on complexity ( patras and olympia , july 2004 ) , ed .
t. bountis , international journal of bifurcation and chaos ( 2005 ) , in press [ cond - mat/0411073 ] ; c. tsallis , m. gell - mann and y. sato , _ special scale - invariant occupancy of phase space makes the entropy @xmath7 additive _ , [ cond - mat/0502274 ] preprint , 2005 . , edited by s. abe and y. okamoto , lecture notes in physics vol . 560 ( springer - verlag , heidelberg , 2001 ) ; _ non - extensive thermodynamics and physical applications _ , edited by g. kaniadakis , m. lissia , and a. rapisarda [ physica a * 305 * ( 2002 ) ] ; _ anomalous distributions , nonlinear dynamics and nonextensivity _ , edited by h. l. swinney and c. tsallis [ physica d * 193 * ( 2004 ) ] .
_ nonextensive entropy - interdisciplinary applications _ , edited by m. gell - mann and c. tsallis ( oxford university press , new york , 2004 ) | the cornerstone of boltzmann - gibbs ( @xmath0 ) statistical mechanics is the boltzmann - gibbs - jaynes - shannon entropy @xmath1 , where @xmath2 is a positive constant and @xmath3 a probability density function .
this theory has exibited , along more than one century , great success in the treatment of systems where short spatio / temporal correlations dominate .
there are , however , anomalous natural and artificial systems that violate the basic requirements for its applicability .
different physical entropies , other than the standard one , appear to be necessary in order to satisfactorily deal with such anomalies .
one of such entropies is @xmath4^q)/(1-q)$ ] ( with @xmath5 ) , where the entropic index @xmath6 is a real parameter .
it has been proposed as the basis for a generalization , referred to as _ nonextensive statistical mechanics _ , of the @xmath0 theory . @xmath7
shares with @xmath8 four remarkable properties , namely _ concavity _ ( @xmath9 ) , _ lesche - stability _ ( @xmath9 ) , _ finiteness of the entropy production per unit time _ ( @xmath10 ) , and _ additivity _ ( for at least a compact support of @xmath6 including @xmath11 ) .
the simultaneous validity of these properties suggests that @xmath7 is appropriate for bridging , at a macroscopic level , with classical thermodynamics itself . in the same natural way that exponential probability functions arise in the standard context
, power - law tailed distributions , even with exponents _ out _ of the lvy range , arise in the nonextensive framework . in this review ,
we intend to show that many processes of interest in economy , for which fat - tailed probability functions are empirically observed , can be described in terms of the statistical mechanisms that underly the nonextensive theory . |
the study of the vortex properties in type - ii superconductors is of extreme interest both for investigating the basic physics of the superconductivity and for evaluating the quality of the materials in view of practical applications .
many aspects of this subject are still topical , especially with regard to the high temperature superconductors ( hts ) and the recently discovered iron - based superconductors . to date , five families of fe - based superconductors have been discovered : _ _ re__ofeas , ( 1111 , _ _ re__=rare earth),@xcite _
_ a__fe@xmath4as@xmath4 ( 122 , _ _ a__=alkaline earth ) , @xcite _ _ x__feas ( 111 , _ _ x__=li ; na),@xcite fe(se , ch ) ( 11 , ch = s , te ) @xcite and the most recently discovered 21311 family of sr@xmath4mo@xmath5fepn ( m = sc , v , cr and pn = pnictogen).@xcite among these families , iron chalcogenides are considered of particular interest because of their simple crystal structure consisting of fe ions tetrahedrally coordinated by se and te arranged in layers stacked along the c - axis , without any other interlayer cations , as occurs in the pnictides .
for this reason , iron chalcogenides are generally considered an ideal candidate for understanding some open issues of high - temperature superconductivity .
one of the most intriguing phenomena observed in the study of the vortex properties in type - ii superconductors is the so called peak ( or fishtail ) effect .
a second peak in the magnetization curve @xmath7(@xmath8 ) of low-@xmath9 superconductors has been observed as early as in the sixties@xcite and a first theoretical explanation has been suggested by pippard in 1969.@xcite the peak has been found in proximity of @xmath10 in nb@xmath5sn , @xcite ceru@xmath4 , @xcite nbse@xmath4 , @xcite v@xmath5si@xcite and mgb@xmath4 .
@xcite + the peak effect has been clearly observed also in the cuprate superconductors but its features vary from one compound to the other . in yba@xmath4cu@xmath5o@xmath6 crystals ,
the peak is broad and temperature dependent ; its origin has been attributed to the presence of inhomogeneities in the oxygen content@xcite and associated to different mechanisms.@xcite in the more anisotropic bi@xmath4sr@xmath4ca@xmath11cu@xmath12o@xmath13 the peak in the magnetization loop has usually been associated to a 3d to 2d transition of the vortex structure .
@xcite however , significant differences in the second - peak features have been observed within this family : in the highly anisotropic bi-2212 and bi-2223 the peak is sharp and temperature independent @xcite while in bi-2201 crystals it exhibits a strong temperature dependence . @xcite regarding the iron - based superconductors ,
the peak effect has been observed in feas-1111 , @xcite in feas-122 @xcite and in fete@xmath14se@xmath15,@xcite even if a clear understanding of the physical origin of the phenomenon in this family has not been achieved , yet .
however , since it has been shown by small angle neutron scattering,@xcite bitter decoration@xcite and magnetic force microscopy@xcite that an ordered vortex lattice is not present in these compounds , the origin of the peak effect due to a phase transition of the vortex lattice seems to be excluded . beyond the investigation of the magnetization curves ,
an useful approach for investigating the vortex dynamics is the study of the relaxation processes of the magnetization .
the critical state of the vortex lattice , which determines the hysteresis of the magnetization in type - ii superconductors , is a metastable state .
it follows that vortices tend to hop out of their pinning - potential well in order to reach the configuration of absolute minimum energy .
such motion usually arises from thermal activation , but it can also arise from quantum tunneling ( at low temperatures ) or can be stimulated by external perturbations , such as microwave shaking of the vortex lattice .
@xcite magnetic relaxation processes have been observed in various low - temperature superconductors .
@xcite however , the subject has become of even greater interest after the discovery of high-@xmath9 superconductors , because of the higher operating temperatures and of the small activation energies related to the short coherence length and large anisotropy .
the concept of thermally - induced hopping of the flux lines has been first treated by anderson and kim .
@xcite in the framework of their model a logarithmic dependence of the magnetization ( @xmath7 ) on the time ( @xmath16 ) is expected .
this behavior has been verified in various superconductors , both low @xmath9 and high @xmath9 ; however , in many experiments , deviations from the logarithmic dependence of @xmath7 have been observed , indicating a failure of the approximations that bring this expectation .
@xcite in the framework of the collective pinning theory a more complex expression for the dependence of the potential energy barrier height ( the so - called interpolation formula ) is proposed : @xmath17 $ ] where @xmath18 is a parameter varying as a function of the vortex - vortex and vortex - pinning center interaction .
@xcite for example , in three dimensional systems @xmath19 when the creep is dominated by the motion of individual flux lines , @xmath20 in the case of collective creep of small bundles and @xmath21 when the bundle size is much larger than the larkin correlated volume @xcite . in 1991 ,
vinokur , feigelman and geshkenbein proposed a theoretical model for the thermally activated flux creep , assuming a logarithmic dependence of the activation energy on the critical current , @xmath22 .
@xcite this dependence of @xmath23 over @xmath24 is a good approximation for the creep activation barrier in the single vortex creep regime ( limit @xmath25 ) and provides a proper fit of the experimental @xmath26 dependence observed in different superconductors .
@xcite the divergence of @xmath26 when @xmath27 can be understood in the context of the collective pinning considering that in the vortex - glass state vortex motion is possible only in the presence of a current .
@xcite in the case of this logarithmic dependence of the activation energy on the critical current , a power - law time dependence of the magnetization is expected .
@xcite in this paper we study the magnetic properties of fete@xmath0se@xmath1 crystals in the superconducting state by investigating the dependence of the magnetic moment ( @xmath28 ) on the applied magnetic field ( @xmath8 ) the temperature ( @xmath29 ) and the relaxation time ( @xmath16 ) . in sec .
[ exp ] we report details on the examined samples and on the experimental techniques used . the experimental results are shown and discussed in two different sections . in sec .
[ vortpin ] , dedicated to the vortex pinning properties , we analyze the @xmath30 curves and deduce the pinning mechanism in the investigated crystals .
the magnetic relaxation processes are studied in sec .
[ vortdyn ] .
conclusions are reported in sec .
[ conclusions ] .
the fete@xmath0se@xmath1 crystals investigated were grown by a modified bridgman - stockbarger method , starting from a nominal fe:(te , se ) ratio of 0.9:1 .
details of the procedure used for preparing the samples are reported in ref .
the refined composition of the samples is fe@xmath31te@xmath32se@xmath33 , indicating a very low iron excess.@xcite it has been shown that reducing the fe excess in fe@xmath34te@xmath35se@xmath36 favors the occurrence of superconductivity and weakens the antiferromagnetic order .
@xcite however , according to the phase diagram of the compound , @xcite a little excess of fe is needed for stabilizing the structure .
it is also worth noting that , due to se - doping for te , less fe is allowed to occupy the additional site , since both the effects of reducing @xmath37 and increasing @xmath38 result in shrinking and re - shaping the fete@xmath39 tetrahedra ; so , the lower is the se content , the more difficult is to limit the fe excess.@xcite two different crystals , weighting @xmath40 mg ( sample a ) and @xmath41 mg ( sample b ) , have been investigated in the present study .
the susceptibility curve @xmath42 of the two samples has been deduced from a zero field cooled @xmath43 measurement performed at 10 oe by a superconducting quantum interference device ( squid ) with the field applied along the @xmath3 axis .
both curves have been corrected for the demagnetization effects according to the formula reported in ref.@xcite .
the results are shown in fig.[chi ] : the @xmath42 curve of sample a exhibits a superconducting transition with @xmath44 k ,
@xmath45 k whilst for sample b it results @xmath46 k , @xmath47 k. both the samples exhibit bulk superconductivity with @xmath48 .
temperature dependence of the dc volume susceptibility ( demagnetization corrected ) for the investigated samples.,width=8 ] the samples have been characterized also by x - ray diffraction ( xrd ) both in a powder diffractometer ( on manually ground crystals ) and in a 4-circle diffractometer .
the rocking curves indicate the presence of a distribution of the c - axis crystallites with a full width at half maximum ( fwhm ) of about @xmath49 for the sample a and about @xmath50 for sample b. our results indicate that , whereas the sample b is of high crystalline quality , the sample a is more likely to be composed of few single - crystalline domains well aligned along the crystallographic c - axis .
the magnetization curves @xmath28(@xmath8 ) of the fete@xmath0se@xmath1 samples were acquired by means of a squid magnetometer and a vsm ( vibrating sample magnetometer ) .
the maximum applied magnetic field was 4.5 t for the squid and 8.8 t for the vsm .
measurements were performed in the range of temperatures from 4.25 k to @xmath9 , with the magnetic field applied both parallel and perpendicularly to the @xmath3 axis .
the vsm measurements were executed with different values of the magnetic field sweep rate ( @xmath51 ) namely 0.05 t / min , 1 t / min and 2 t / min .
furthermore , by means of the vsm , we studied the relaxation of magnetization over periods of time up to 6000 sec , for fixed values of temperature and magnetic field . in order to perform measurements in the trapped flux configuration
, the relaxation measurements were performed with the following procedure : i ) the sample is zero - field cooled ; ii ) the magnetic field is increased from zero to 8.8 t with @xmath51= 2t / min ; iii ) the field is decreased from 8.8 t to 8 t with @xmath51= 0.1 t / min ; iv ) the magnetic relaxation at 8 t is recorded for 6000 sec ; iv ) the field is decreased by 1 t with @xmath51= 0.1 t / min ; v ) the magnetic relaxation is recorded for 6000 sec . steps iv ) and v ) are repeated for registering the relaxation down to 0 t , in steps of 1 t.
fig . [ m(h ) ] shows the magnetic - moment curves of the sample a acquired with the vsm at different temperatures , for @xmath52 ( a ) and @xmath53 ( b ) , for a sweep rate of 2 t / min .
increasing the temperature , the pinning becomes weaker and consequently the width of the hysteresis loop decreases .
a second peak in the @xmath7(@xmath8 ) curve is clearly observed for @xmath52 , up to temperatures close to @xmath9 .
a second peak is also present for @xmath53 but less pronounced and detectable only for temperatures higher than 5 k. in the inset of fig .
[ m(h ) ] the results obtained for the sample b at @xmath54 k and @xmath55 k , are reported for the two different orientations of the field . in this case , a second peak in the @xmath28(@xmath8 ) curve is not observed for @xmath53 , whereas it is clearly evident for @xmath52 , in analogy to what observed in yba@xmath4cu@xmath5o@xmath6 crystals@xcite and in other fe - based superconductors.@xcite since sample a presents a wider c - axis distribution of single - crystalline domains than sample b , we argue that the weak peak effect revealed for @xmath53 in this sample is due to the slight misalignment of the crystallites forming the sample .
we would also like to specify that the experimental @xmath30 curves of sample b for @xmath52 are qualitatively similar with those of sample a at comparable temperatures .
this suggests that a misalignment of the layers forming the crystal is crucial only for magnetic fields applied perpendicularly to the @xmath3 axis of the sample .
the @xmath30 curves obtained for @xmath56 exhibit a paramagnetic - like background arising from the experimental setup .
this background is not observed in the @xmath30 curves measured for @xmath56 in the squid . in any case , the background does not alter the results obtained in the following , which rely only on the separation between the positive and negative branches of the magnetic - moment loop .
magnetic moment vs. external field curves obtained at different temperatures in sample a for @xmath52 ( a ) and @xmath56 ( b ) .
the insets show the results obtained in sample b at @xmath57 k and @xmath55 k.,width=8 ] from the @xmath28(@xmath8 ) curves measured for @xmath52 at different @xmath29 , we determined the temperature dependence of the magnetic field value corresponding to the second peak in magnetization ( @xmath58 ) . the results for samples a and b are shown in fig .
[ h - peak(t ) ] .
it follows that for the fete@xmath0se@xmath1 superconductor the position of the second - peak shifts towards lower fields monotonically on increasing the temperature , in analogy to what observed in the yba@xmath4cu@xmath5o@xmath6 superconductor .
@xcite for both samples investigated , the peak becomes undetectable at @xmath59 k. the continuous lines in fig .
[ h - peak(t ) ] are the best fit curves obtained supposing the same @xmath29-dependence of @xmath58 as observed in yba@xmath4cu@xmath5o@xmath6:@xcite @xmath60 , where @xmath61 is a constant and @xmath62 is the temperature at which the peak is undetectable .
the best fit parameters ( with the relative statistical errors ) are : @xmath63 t , @xmath64 k for sample a and @xmath65 t , @xmath66 k for sample b. this remarkable similarity between the results obtained in the yba@xmath4cu@xmath5o@xmath6 and in the fete@xmath0se@xmath1 superconductors , might suggest that the peak effect in the two systems has an analogous origin .
temperature dependence of the magnetic field value at which the peak effect occurs , for the two measured samples .
continuous lines are the best fit curves obtained supposing the same @xmath29 dependence observed in yba@xmath4cu@xmath5o@xmath6.,width=8 ] the magnetic field dependence of the critical current density ( @xmath67 ) has been extracted from the @xmath28(@xmath8 ) curves , for different values of the temperature , using the bean critical state formulae.@xcite for slab samples in a perpendicular magnetic field : @xmath68 , where @xmath69(@xmath70 ) is the separation between the two branches of the magnetic - moment loop , @xmath71 and @xmath72 are the length and the width of the sample ( @xmath73 ) and @xmath74 is the thickness .
the @xmath67(@xmath8 ) curves obtained for the sample a at different temperatures are presented in fig .
[ jc ] ; in the inset of the same figure the @xmath67(@xmath8 ) curve at @xmath57 k of the sample b is shown . for both samples investigated the @xmath67 values at 4.25 k are of the order of @xmath75 a/@xmath76 .
this value is smaller than what measured in fe@xmath34te@xmath35se@xmath36 crystals with higher @xmath38 values than our samples.@xcite however , the @xmath67 values become comparable if related to the same reduced temperature @xmath77.@xcite the fact that the @xmath67 values obtained for sample a are very similar to those obtained for the high - quality single crystal sample b , where no weak links are present , indicates that weak links effects are not important for sample a either . the critical current density @xmath67 is always limited by the depairing current density @xmath78 , where @xmath79 and @xmath80 are the thermodynamic critical field and the gl penetration depth , respectively .
@xcite important information can be deduced from the ratio @xmath81 . for low-@xmath9 superconductors pinning
is usually strong ( @xmath82 ) resulting from the interaction of vortices with extended defects such as , e.g. , precipitates or grain boundaries.@xcite on the contrary , for the cuprate family pinning is usually weak ( @xmath83 ) normally arising from point defects , e.g. , oxygen vacancies.@xcite for the fete@xmath0se@xmath1 we estimate @xmath84 indicating that also in the case of the fe - based 11 family pinning is weak . the microscopic origin of pinning in fe@xmath34te@xmath35se@xmath36 will be discussed in the following .
field dependence of the critical current density at different temperatures and for @xmath85 , in sample a. the inset shows the results obtained at @xmath57 k and for @xmath85 for the sample b.,width=8 ] as already mentioned , the presence of the peak effect in the magnetization curve has been widely documented in the literature , both in low and high @xmath9 superconductors .
its origin has been associated to different processes , depending on the particular system investigated .
@xcite in order to shed light on the mechanisms that rule pinning in the fe@xmath34te@xmath35se@xmath36 superconductor , we have investigated the magnetic - field dependence of the pinning force density , @xmath86 .
it has been shown for a large variety of low-@xmath9 and high-@xmath9 superconductors that the curves of @xmath86 vs. @xmath8 obtained at different temperatures may be scaled into a unique curve if they are plotted as a function of the reduced field , @xmath87.@xcite the empirical formula that accounts for the scaling is : @xmath88 where @xmath89 is a proportionality constant , @xmath90 and @xmath91 are two parameters whose values depend on the origin of the pinning mechanism .
@xcite the different contributions to flux - pinning are usually catalogued into two main categories : i ) @xmath2 ( or normal ) pinning , arising from spatial variations in the charge carrier mean free path ( @xmath71 ) ; ii ) @xmath92 ( or @xmath93 ) pinning , associated with spatial variations of the ginzburg parameter ( @xmath94 ) due to fluctuations in the transition temperature @xmath9 .
@xcite in the last category fall
also non - superconducting metallic particles whose dimension is smaller than the coherence length ( @xmath95 ) since , in this case , the superconductivity is induced by the proximity effect .
a classification is also made for pinning centers , as a function of the number of dimensions that are large with respect to the inter - vortex distance @xmath96 .
following the definition given by dew - hughes in ref.@xcite , in this paragraph we refer to point pins as regions whose dimensions in all directions are less than @xmath74 , line pins have one dimension larger than @xmath74 , grain- and twin- boundaries , which have two dimensions greater than @xmath74 , act as surface pins , volume pins have all dimensions large with respect to @xmath74.@xcite in the framework of the dew - hughes model,@xcite different values for @xmath90 and @xmath91 in eq .
[ fp ] are expected , as a function of the specific pinning mechanism involved .
correspondingly , the theoretical @xmath86 vs. @xmath97 curves present a maximum at different @xmath97 values . in the case of @xmath2 pinning ,
the maximum is expected at @xmath98 ( @xmath99 , @xmath100 ) for point pins and at @xmath101 ( @xmath102 , @xmath100 ) for surface pins , such as grain boundaries ( the same dependence has been predicted for shear - breaking in the case of a set of planar pins @xcite ) ; no maximum is expected in the case of @xmath2 volume pinning , being @xmath103 and @xmath100 .
the maximum of the @xmath104 curve is expected at higher @xmath97 values in the case of @xmath92 pinning ; in particular it occurs at @xmath105 ( @xmath106 , @xmath107 ) for point pins , at @xmath108 ( @xmath109 , @xmath107 ) for surface pins and at @xmath110 ( @xmath99 , @xmath107 ) for volume pins .
therefore , important information on the physical origin of the pinning mechanisms can be achieved by analyzing the scaled @xmath104 curves .
the comparison of the predicted pinning functions with the experiments requires an estimation of @xmath111 , defined as the @xmath8-value at which @xmath112 . starting from the experimental @xmath113 curves ,
@xmath111 is usually determined as the extrapolated zero value in the so - called kramer plot,@xcite where @xmath114 is plotted as a function of @xmath8 .
this procedure has been successfully used in the case of wires , polycrystalline samples and crystals.@xcite however , if pinning is ruled only by defects whose dimensions are smaller than the intervortex distance , it is expected @xmath115.@xcite in this case , @xmath111 may be determined by doing a linear extrapolation down to zero of the @xmath116 vs. @xmath8 curve.@xcite in fig .
[ kramerplot ] ( a ) and ( b ) we report the kramer plots along with the @xmath116 vs. @xmath8 curves , obtained at two different temperatures , for sample a and sample b , respectively .
we observe that the kramer plot presents a wide linear behavior for both samples . on the other hand ,
the @xmath116 curve exhibits a linear behavior in a wide range of magnetic fields , only for sample b. for this reason , we have extracted @xmath111 from the kramer plot for sample a , whereas for sample b we can not discriminate a - priori which is the most correct procedure .
however , a more conclusive result can be achieved if one considers simultaneously both the @xmath67 vs. @xmath8 dependence and the scaled @xmath117 curves .
if one hypothesizes that pinning is ruled by point pins and consequently extracts @xmath111 from the @xmath116 vs. @xmath8 curve then the corresponding @xmath117 curve should exhibit a maximum at @xmath118 .
kramer plot ( right axis ) and @xmath116 vs. @xmath8 ( left axis ) curves obtained at @xmath119 k and @xmath120 k for sample a ( a ) and sample b ( b ) , as described in the text.,width=8 ] fig .
[ fpinning ] ( a ) and ( b ) present the normalized pinning - force density for samples a and b at different temperatures as a function of the reduced field based on @xmath111 values deduced from the kramer plots , whereas the inset of fig . 6 ( b )
is based on @xmath111 values determined from a linear @xmath116 vs. @xmath8 dependence .
for the sake of clearness , the superscript of @xmath121 and @xmath122 in the figure indicate respectively the @xmath111 values extracted from the kramer plot or hypothesizing @xmath115 .
normalized pinning force density curves as a function of the reduced field , @xmath123 , for sample a ( a ) and b ( b ) .
@xmath121 indicates that the irreversibilty field has been deduced by a linear extrapolation in the kramer plot .
the inset in fig .
[ fpinning ] ( b ) shows the results obtained in sample b , scaled by using the irreversibility field , @xmath122 , deduced by the @xmath116 vs. @xmath8 curves , width=8 ] the pinning force curves of sample a , obtained at different temperatures , scale well and present a maximum at @xmath124 while , for sample b , the maximum occurs at @xmath125 if one considers @xmath126 or at @xmath118 if @xmath127 .
these results indicate that , for the single crystal sample b , the pinning mechanisms is ruled by defects whose dimensions are smaller that the intervortex distance in the investigated field range and as a consequence , the most appropriate procedure for determining @xmath128 is from the @xmath116 vs. @xmath8 curve . on the contrary , for sample
a the position of the maximum in the @xmath117 curve suggests that a contribution to pinning coming from surface pinning can not be excluded .
from the scaled pinning - force density curves we can also deduce information on the type of pinning centers . in particular , the fact that the maximum in the @xmath104 curve occurs for values of the reduced field @xmath129 is an indication that the pinning centers in fete@xmath0se@xmath1 are of @xmath2 type .
it is worth noting that this conclusion does not depend on the specific procedure used for determining h@xmath130 .
in fact , in the case of pinning contribution coming from @xmath92-type pins , the maximum is expected to occur at @xmath131 , @xcite which is far away from what observed in our samples . the analysis of the scaled pinning force curves in the frame of the dew - hughes model revealed to give important information about pinning in low-@xmath9 , cuprates as well as fe - based superconductors.@xcite .
however since in our samples pinning is weak ( @xmath84 ) and thermal fluctuations are large ( ginzburg number @xmath132 ) , more conclusive results about pinning origin in fete@xmath0se@xmath1 are expected to be achieved by analyzing the experimental results in the framework of the collective pinning theory.@xcite following the theoretical approach proposed by griessen _
et al _ , @xcite in the case of @xmath2-type weak pinning in the single vortex regime it is expected that the critical - current density variation with respect to the reduced temperature ( @xmath133 ) is described by the following expression:@xcite @xmath134 while , for @xmath92 pinning , it is : @xmath135 in fig .
[ griessengraph ] , we plot the normalized @xmath136 data obtained at @xmath137 t and @xmath138 t for @xmath85 , along with the theoretical curves expected within the scenario of @xmath2 and @xmath92 pinning .
the @xmath136 values have been extracted from the @xmath67(@xmath8 ) curves obtained at various temperatures .
analogous results have been obtained for sample b. a remarkably good agreement between the experimental results and the @xmath2 pinning theoretical curve is obtained , confirming that pinning in the fete@xmath0se@xmath1 samples originates form spatial variation of the mean free path .
the weak collective pinning theory considers that pinning originates from fluctuation in the density and force of defects whose dimension is smaller than the coherence length . in this case
small values for the critical current density are expected ( @xmath139).@xcite on the contrary , if pinning was mainly ruled by extended defects such as , e.g. , precipitates , twinning or grain boundaries , or columnar defects such as dislocation lines , higher values for the @xmath81 ratio would be expected ( @xmath140).@xcite a phase separation in fe@xmath34te@xmath35se@xmath36 single crystals with different @xmath37 and @xmath38 values has been experimentally observed by scanning tunneling microscopy and electron energy - loss spectroscopy .
@xcite in particular , fluctuations of the local te concentration ( even by 20@xmath141 from the average local composition ) in nanometric regions have been observed in fete@xmath0se@xmath1 crystals .
@xcite these fluctuations could be the origin of the observed weak pinning . in spite of the inhomogeneous chemical distribution ,
our results indicate that local fluctuations of the critical temperature are not important for pinning .
this interesting result is in agreement with tunneling spectroscopy experiments performed at @xmath142 k in optimal doped fete@xmath143se@xmath144 crystals , indicating that the compound is chemically inhomogeneous but electronically homogeneous.@xcite normalized critical current density data , as a function of the reduced temperature ( @xmath145 ) , obtained at @xmath146 t and @xmath147 t for @xmath52 .
the continuous lines are the theoretical curves expected in the case of @xmath2 and @xmath92 pinning.,width=8 ] it is worth mentioning that a coexistence of superconducting and magnetic orders has been proposed for the fe@xmath34te@xmath35se@xmath36 system .
in particular such coexistence has been inferred by khasanov _
based on muon - spin rotation experiments.@xcite on the other hand , li _ et al .
_ have ascribed the apparent coexistence of superconductivity and magnetism in the fe@xmath34te@xmath35se@xmath36 system to a phase separation in the real space.@xcite as already mentioned , a phase separation has also been observed in fete@xmath143se@xmath144 crystals by stm.@xcite other authors have assumed a spin - glass state to exist between the long range antiferromagnetic order ( at low se content ) and the bulk superconducting state , at high se content.@xcite the low - field dc susceptibility measurements of our samples , reported in fig.[chi ] , do not show any anomaly or feature attributable to the coexistence of superconductivity and magnetic order . in a previous article
we have deeply investigated the effect of the actual chemical composition and its effect on the crystal chemistry and on the magnetic and superconducting phase diagram.@xcite we have proved the combined effect of fe - excess ( @xmath37 ) and se - substitutions ( @xmath38 ) on the transition from a superconducting to an antiferromagnetic state and highlighted the importance of controlling the real composition in a three dimensional phase diagram . by keeping under control a low fe - excess , we have obtained bulk superconductivity at se - content even lower than @xmath148.@xcite in 2010 , bendele _ _ et al.__@xcite has confirmed our study on the excess fe and merged data from previous publications in a three dimensional phase diagram , still claiming a coexistence of superconductivity and magnetism . on the basis of our experimental results and the contradicting results reported in the literature ,
it is not possible to discriminate whether such coexistence occurs intrinsically at the atomic level in fe@xmath34te@xmath35se@xmath36 or a phase separation and local composition fluctuations are responsible for the magnetic behavior observed by some authors .
further studies on very clean and chemically homogeneous samples , carried out with different experimental techniques , are needed in order to clarify this point . at the end of our analysis on the pinning properties we would like to remind that both the dew - hughes and the griessen models have been developed in the single - vortex pinning regime , i.e. neglecting inter - vortex interactions
the validity of this approximation in our case is confirmed by the magnetic relaxation study reported in the next section . in this section
we report a study of the dynamical properties of vortices in the fete@xmath0se@xmath1 superconductor performed by magnetic relaxation measurements .
the analysis has been limited to sample a , since the measured magnetic moment of sample b was too small for achieving a good resolution in a wide range of times and temperatures .
[ figrelax ] shows the relaxation of the magnetization ( @xmath149 ) normalized to its maximum value , obtained at @xmath57 k for @xmath52 , plotted in a log - log graph .
at all the fields investigated and for times greater than about 50 sec , a linear dependence of @xmath7 has been observed , indicating a power - law dependence of @xmath7 vs. @xmath16 .
an analogous behavior has also been detected at higher temperatures and for @xmath53 .
it has been shown that deviations from the expected @xmath150 curves at short times could be due to the magnetic field overshoot occurring when the external field ramp is stopped.@xcite this overshoot produces a shielded flux zone in proximity of the surface of the sample , which affects the initial relaxation process .
the inset of fig .
[ figrelax ] shows the @xmath151d @xmath152d @xmath153 values obtained at different @xmath8 values ; large values of @xmath154 have already been observed in the fe@xmath34te@xmath35se@xmath36 system as well as in other fe - based superconductors.@xcite since the @xmath154 value is related to the pinning potential energy barrier height , @xcite the minimum observed in the @xmath154(@xmath8 ) curve is associated to the presence of the peak effect in the @xmath67(@xmath155 ) curves .
plot of the magnetization vs. time ( on a log - log scale ) at @xmath57 k for different values of the magnetic field .
the inset shows the field dependence of the relaxation rate @xmath151d @xmath152d @xmath153.,width=8 ] as already described in the introduction , in the framework of the anderson and kim theory , a linear dependence of @xmath7 on log(@xmath16 ) is expected ; this result stems from two basic assumption : i ) the pinning potential energy barrier height decreases linearly with the current density : @xmath156 ; ii ) @xmath157 , which allows to hypothesize that the thermal - induced hopping rate is proportional to the arrhenius factor @xmath158 .
supposing a linear dependence of @xmath23 on @xmath24 is only a first - order approximation whose validity has been demonstrated to fail many times .
@xcite as reported by vinokur , feigelman and geshkenbein , a power - law dependence of @xmath7 vs. @xmath16 is expected if a logarithmic dependence of the activation energy on the current density is supposed : @xmath159 .
@xcite in this case , the following expressions for the time dependence of @xmath7 are predicted : @xmath160(k_bt
/ u_0)\ln(t/\tau_0 ) \ , , \textrm{for}~ t\ll t^*\ ] ] @xmath161 where @xmath74 is the thickness of the sample , @xmath162 is the position of the flux front and @xmath163 is the time at which the sample is fully penetrated .
@xcite in order to verify that a logarithmic dependence of @xmath23(@xmath24 ) is actually present for the fete@xmath0se@xmath1 crystal investigated , we used the method proposed by maley _ et al._,@xcite which allows the determination of the @xmath23(@xmath24 ) curve from the experimental data obtained at different temperatures . in their paper
, the authors show that choosing a proper value for a time - independent constant ( @xmath61 ) it is possible to deduce the @xmath24 dependence of @xmath23 ( or equivalently the @xmath23(@xmath7 ) dependence ) by plotting @xmath164d@xmath165d@xmath166 $ ] vs. @xmath167 , where m@xmath168 is the magnetization at the equilibrium .
this procedure is valid under the assumption that the temperature dependence of @xmath23 is weak and that the principal effect of increasing the temperature is to produce monotonically decreasing initial values of @xmath7 , which is usually valid for @xmath169 .
@xcite fig .
[ figmaley ] shows the curves obtained at @xmath170 t , scaled considering a=28 , along with the best fit curve deduced supposing the following dependence of @xmath23 on @xmath7 : @xmath171 ; the best fit parameters ( with the relative statistical errors ) are : @xmath172 k , @xmath173 emu/@xmath174 and @xmath175 emu/@xmath174 .
similar values for @xmath176 have been obtained in other fe - based superconductors , indicating that the weakness of pinning is a general characteristic of these compounds.@xcite the negative value of @xmath177 is due to the diamagnetic contribution of the sample holder . a very good scaling
is obtained for @xmath178 k , while at @xmath119 k a discrepancy between the experimental data and the theoretical curve starts to be present . as already mentioned in the introduction ,
the logarithmic dependence of @xmath23 on @xmath24 is a good approximation for the creep activation barrier in the single vortex creep regime . as a consequence ,
our results indicate that in the fete@xmath0se@xmath1 superconductor the vortex motion develops in the single - vortex pinning limit even in magnetic fields up to 8 t. this is in general unexpected . at high fields ,
the intervortex distance becomes small compared to the magnetic field penetration depth and thus one would expect that vortex - vortex interactions become important and pinning involves vortex bundles .
however , single vortex pinning up to 9 t has already been observed by inosov _
_ in bafe@xmath179co@xmath15as@xmath4 single crystals by magnetization measurements , small - angle neutron scattering , and magnetic force microscopy.@xcite our results suggests that also in the fe@xmath34te@xmath35se@xmath36 system the vortex - defect interaction dominates up to high fields .
this experimental evidence further confirms the validity of our study on the pinning properties as reported in sec .
[ vortpin ] , carried out in the framework of the dew - hughes model where flux - lattice elasticity effects are neglected .
magnetization dependence of the pinning potential energy barrier height calculated in the frame of the maley model , scaling the data at different temperatures as described in the text.,width=8 ] in sec [ vortpin ] we have shown that the @xmath30 curves obtained for @xmath52 are characterized by a second peak whose position changes with the temperature . in order to investigate the magnetic relaxation effects on the peak effect , we measured the field dependence of the magnetic moment , by means of the vsm , for different sweep rates ( @xmath51 ) of the magnetic field .
it has been demonstrated that the dependence of the hysteresis amplitude on the sweep rate contains basically the same information of the time dependence of the magnetization during relaxation , in particular , the higher is the sweep rate , the shorter is the effective `` observing time '' in the @xmath28(@xmath16 ) curve.@xcite the results are shown in fig .
[ figpeakposition ] , along with the curve acquired by the squid and the values ( full points ) extracted from the @xmath180 curves , obtained at different fields , considering @xmath181 sec . regarding the measurement performed with the squid
, one can assume @xmath182 since any variation of @xmath8 during the acquisition can be neglected .
furthermore , since the acquisition time is of some seconds , one has also to consider that the measured @xmath28 value has already relaxed .
a remarkable variation of the @xmath30 curves associated to different characteristic times is present , as a consequence of the relaxation in time of the magnetic moment .
furthermore , it is very interesting to note that the position of the second - peak moves towards low @xmath8 values during the relaxation , which indicates that the magnetic field value at which the second peak in the @xmath113 curve is obtained relaxes in time .
time evolution of the second peak position investigated by magnetization measurements performed by the vsm at different sweep rates ( @xmath51 ) , by the squid as well as by magnetic relaxation measurements.,width=8 ] in order to shed light on the mechanism that determines the temperature and time dependence of the peak position , @xmath58 , we collected the couples of values ( @xmath58 , @xmath183 ) which identify its location in the @xmath28(@xmath8 ) curves obtained at different temperatures by the squid and by the vsm and for different sweep rates , the results being shown in fig . [ hpeak - mpeak ] .
the inset at the bottom right of the same figure presents the calculated induction - field values ( @xmath184 ) corresponding to @xmath183 .
it is very interesting to note that the pairs of values ( @xmath58 , @xmath183 ) - or equivalently ( @xmath184 , @xmath183 ) - obtained at different temperatures or sweep rates lie on the same curve and that , in some cases , the same pair of values ( @xmath58 , @xmath183 ) has been obtained at two different temperatures but for a different `` observing time '' during relaxation .
the continuous line shown in the figure is the best fit curve obtained by assuming @xmath185 .
the best fit parameters , along with the corresponding statistical errors , are : @xmath186 emu / t ; @xmath187 .
pairs of values ( @xmath58,@xmath183 ) that identify the second peak position in the @xmath30 curve at different temperatures and characteristic times .
labels 1 , 2 , 3 , indicate the measurements performed by the vsm with @xmath51= 2t / min , 1t / min and 0.05 t / min , respectively ; label 4 indicates the measurements performed by the squid .
the continuous line is the best fit curve obtained as described in the text .
the inset at the bottom right shows the couples of values ( @xmath184,@xmath183 ) corresponding to the peak effect .
the inset at the top left is an imaginary drawing of the vortex energy landscape , the line indicating the relaxation path of the vortex configuration associated to the peak effect.,width=8 ] since the induction field @xmath155 is proportional to the number of vortices present in the sc and the magnetic moment identifies the critical current ( and as a consequence the flux profile inside the sample ) , the pair of values ( @xmath184 , @xmath183 ) identify a specific configuration of the vortex structure .
the fact that all the experimental data shown in fig .
[ hpeak - mpeak ] fall in a unique curve suggests that there is a unique vortex configuration which determines the occurrence of the peak effect at a given @xmath58 value , whatever the temperature is in the range 4.2 k @xmath188 8 k , corresponding to 0.4 @xmath189 0.75 @xmath9 . in particular , the same ( @xmath184,@xmath183 ) pair can be found at a temperature @xmath190 and observing time @xmath191 or equivalently at a temperature @xmath192 at a time @xmath193 . if one defines the vortex energy landscape as the mapping of all the possible configurations of the vortices in the sample , determined by @xmath155 and @xmath28 , and the corresponding energy level ( @xmath194 ) , the observed behavior could be justified by supposing that there is a unique path in the vortex energy landscape that describes the relaxation of the vortex configuration associated to the peak effect and that the temperature does not particularly affect the energy landscape in the @xmath29 range explored , at least in proximity of the path .
the only effect of increasing the temperature is to allow the relaxation to start from a point in the path closer to the final equilibrium state .
this result also indicates that the mechanism behind the second peak is related to a thermally driven process in the examined range of temperatures and fields .
we are extending the present analysis of the peak effect properties to other superconductors , such as yba@xmath4cu@xmath5o@xmath6 , and the results will be discussed in a forthcoming paper .
we have investigated the magnetic properties of fete@xmath0se@xmath1 crystals in the superconducting state by magnetization and magnetic relaxation measurements .
we have shown that pinning in fete@xmath0se@xmath1 originates form spatial variation of the mean free path and is most likely related to the nanoscale chemical phase separation observed in fe@xmath34te@xmath35se@xmath36 on a scale of @xmath195 nm.@xcite very interestingly , our results confirm that even if chemically inhomogenous fe@xmath34te@xmath35se@xmath36 is electronically homogeneous since pinning is not ruled by spatial fluctuations of the critical temperature . from magnetic relaxation measurements
we have obtained indications that vortex motion develops in the single vortex limit even in magnetic field as high as 8 t , in agreement to what observed in the 122 fe - based compounds.@xciteon applying the magnetic field along the @xmath3 axis , a clear peak effect in the @xmath30 curves has been observed , up to temperatures near @xmath9 .
the second - peak position varies with the temperature , following the same dependence as observed in yba@xmath4cu@xmath5o@xmath6.@xcite the relaxation of the vortex configuration that determines the peak effect has also been studied .
we have found that the pairs of values ( @xmath58 , @xmath183 ) that identify the second - peak position in the @xmath30 curves obtained at different temperatures and different relaxation times reconstruct a unique curve .
this suggests that the vortex configuration that determines the peak effect at a particular @xmath8 value does not depend on the temperature in the range 0.4 @xmath9 @xmath188 0.75 @xmath9 .
it follows that the temperature influence on the vortex - vortex and vortex - defect interactions in proximity of the peak effect is negligible .
we acknowledge prof . thierry giamarchi and dr .
christophe berthod for useful discussions , dr . florin buta for critical reading of the paper , damien zurmuehle and alexandre ferreira for technical support .
99 y. kamihara , t. watanabe , m. hirano , h. hosono , j. am .
* 130 * ( 2008 )
3296 - 3297 . m. rotter , m. tegel , d. johrendt , i. schellenberg , w. hermes , r. pttgen , phys .
b * 78 * , 020503 ( 2008 ) . x. c. wang , q. q. liu , y. x. lv , w. b. gao , l. x. yang , r. c. yu , f. y. li , c. q. jin , solid state communications * 148 * , 538 ( 2008 ) .
d. r. parker , m. j. pitcher , p. j. baker , i. franke , t. lancaster , s. j. blundell , s. j. clarke , chem . commun . * 16 * , 2189 ( 2009 ) .
fong - chi hsu , jiu - yong luo , kuo - wei yeh , ta - kun chen , tzu - wen huang , phillip m. wu , yong - chi lee , yi - lin huang , yan - yi chu , der - chung yan , and maw - kuen wu , proc .
usa * 105 * , 14262 ( 2008 ) .
kuo - wei yeh , tzu - wen huang , yi - lin huang , ta - kun chen , fong - chi hsu , phillip m. wu , yong - chi lee , yan - yi chu , chi - lian chen , jiu - yong luo , der - chung yan and maw - kuen wu , eur .
* 84 * , 37002 ( 2008 ) .
h. ogino , y. matsumura , y. katsura , k. ushiyama , s. horii , k. kishio , j. shimoyama , supercond .
* 22 * , 075008 ( 2009 ) . w. de sorbo , rev .
phys . * 36 * , 90 ( 1964 ) .
a. b. pippard , phil .
* 19 * , 217 ( 1969 ) .
r. lortz , n. musolino , y. wang , a. junod , n. toyota , phys .
b * 75 * , 094503 ( 2007 ) . s. s. banerjee , n. g. patil , s. saha , s. ramakrishnan , a. k. grover , s. bhattacharya , g. ravikumar , p. k. mishra , t. v. chandrasekhar rao , v. c. sahni , m. j. higgins , e. yamamoto , y. haga , m. hedo , y. inada , y. onuki , phys . rev .
b * 58 * , 995 ( 1998 ) . m. isino , t. kobayashi , n. toyota , t. fukase , and y. muto , phys . rev .
b * 38 * , 4457 ( 1988 ) .
m. pissas , s. lee , a. yamamoto , and s. tajima , phys .
lett . * 89 * , 097002 ( 2002 ) .
m. angst , r. puzniak , a. wisniewski , j. jun , s. m. kazakov , and j. karpinski , phys .
b * 67 * , 012502 ( 2003 ) .
l. klein , e. r. yacoby , y. yeshurun , a. erb , g. mller - vogt , v. breit , h. whl , phys . rev .
b * 49 * , 4403 ( r ) ( 1994 ) .
hai - hu wen and zhong - xian zhao , appl .
* 68 * , 856 ( 1996 ) .
v. tobos , l. m. paulius , a. m. petrean , s. ferguson , j. w. snyder , r. j. olsson , w .- k .
kwok , and g. w. crabtree , appl .
* 78 * , 3097 ( 2001 ) .
l. krusin - elbaum , l. civale , v. m. vinokur , f. holtzberg , phys .
lett . * 69 * , 2280 ( 1992 ) .
l. civale , l. krusin - elbaum , j. r. thompson , f. holtzberg , f. , phys .
b * 50 * , 7188 ( 1994 ) .
y. abulafia , a. shaulov , y. wolfus , r. prozorov , l. burlachkov , y. yeshurun , d. majer , e. zeldov , h. whl , v. b. geshkenbein , v. m. vinokur , phys . rev
. lett . * 77 * , 1596 ( 1996 ) .
s. kokkaliaris , p. a. j. de groot , s. n. gordeev , a. a. zhukov , r. gagnon , l. taillefer , phys .
lett . * 82 * , 5116 ( 1999 ) .
b. rosenstein , a. knigavko , phys .
lett . * 83 * , 844 ( 1999 ) .
g. yang , p. shang , s. d. sutton , i. p. jones , j. s. abell , c. e. gough , phys .
b * 48 * , 4054 ( 1993 ) . s. l. lee , p. zimmermann , h. keller , m. warden , i. m. savi , r. schauwecker , d. zech , r. cubitt , e. m. forgan , p. h. kes , t. w. li , a. a. menovsky , z. tarnawski , phys .
* 71 * , 3862 ( 1993 ) .
b. khaykovich , e. zeldov , d. majer , t. w. li , p. h. kes , m. konczykowski , phys .
lett . * 76 * , 2555 ( 1996 ) .
a. piriou and y. fasano and e. giannini and .
fischer , physica c * 460 - 462 * , 408 ( 2007 ) .
a. piriou , y. fasano , e. giannini , .
fischer , phys .
b * 77 * , 184508 ( 2008 ) . t. tamegai and y. iye and i. oguro and k. kishio , physica c * 213 * , 33 ( 1993 ) . n. avraham , b. khaykovich , y. myasoedov , m. rappaport , h. shtrikman , d. e. feldman , t. tamegai , p. h. kes , ming li , m. konczykowski , k. van der beek , e. zeldov , nature * 411 * , 451 ( 2001 ) .
a. piriou , e. giannini , y. fasano , c. senatore , .
fischer , phys .
b * 81 * , 144517 ( 2010 ) .
j. d. moore , k. morrison , k. a. yates , a. d. caplin , y. yeshurun , l. f. cohen , j. m. perkins , c. m. mcgilvery , d. w. mccomb , z. a. ren , j. yang , w. lu , x. l. dong , and z. x. zhao , supercond .
* 21 * , 092004 ( 2008 ) . c. senatore , r. flkiger , m. cantoni , g. wu , r. h. liu , and x. h. chen , phys
b * 78 * , 054514 ( 2008 ) .
r. prozorov , m. e. tillman , e. d. mun and p. c. canfield , new j. phys .
* 11 * , 035004 ( 2009 ) .
huan yang , huiqian luo , zhaosheng wang , and hai - hu wen , appl .
. lett . * 93 * , 142506 ( 2008 ) .
r. prozorov , n. ni , m. a. tanatar , v. g. kogan , r. t. gordon , c. martin , e. c. blomberg , p. prommapan , j. q. yan , s. l. budko , p. c. canfield , phys . rev .
b * 78 * , 224506 ( 2008 ) .
r. prozorov , m. a. tanatar , p. prommapan , r. t. gordon , n. ni , s. l. budko , p. c. canfield , physica c * 469 * , 667 - 673 ( 2009 ) .
bing shen , peng cheng , zhaosheng wang , lei fang , cong ren , lei shan , and hai - hu wen , phys .
b * 81 * , 014503 ( 2010 ) .
s. salem - sugui , jr . ,
l. ghivelder , a. d. alvarenga , l. f. cohen , k. a. yates , k. morrison , j. l. pimentel , jr . ,
huiqian luo , zhaosheng wang , and hai - hu wen , phys .
b * 82 * , 054513 ( 2010 ) .
p. das , ajay d. thakur , anil k. yadav , c. v. tomy , m. r. lees , g. balakrishnan , s. ramakrishnan , a. k. grover , phys .
b * 84 * , 214526 ( 2011 ) . m. r. eskildsen , l. ya .
vinnikov , t. d. blasius , i. s. veshchunov , t. m. artemova , j. m. densmore , c. d. dewhurst , n. and ni , a. kreyssig , s. l. budko , p. c. canfield , a. i. goldman , phys .
b * 79 * , 100501(r ) ( 2009 ) .
d. s. inosov , t. shapoval , v. neu , u. wolff , j. s. white , s. haindl , j. t. park , d. l. sun , c. t. lin , e. m. forgan , m. s. viazovska , j. h. kim , m. laver , k. nenkov , o. khvostikova , s. khnemann , v. hinkov , phys .
b * 81 * , 014513 ( 2010 ) .
m. r. eskildsen , e. m. forgan , h. kawano - furukawa , rep .
. phys . * 74 * , 124504 ( 2011 ) .
g. p. mikitik and e. h. brandt , journal of low temperature physics * 139 * , 221 ( 2005 ) .
y. yeshurun , a. p. malozemoff , a. shaulov , rev .
physics * 68 * , 911 ( 1996 ) .
p. w. anderson , phys .
* 9 * , 309 ( 1962 ) .
p. w. anderson , y. b. kim , rev .
* 36 * , 39 ( 1964 ) .
d. zola , m. polichetti , c. senatore , and s. pace , phys .
b * 70 * , 224504 ( 2004 ) . m. v. feigelman , v. b. geshkenbein , a. i. larkin , and v. m. vinokur , phys .
lett . * 63 * , 2303 ( 1989 ) .
g. blatter , m. v. feigelman , v. b. geshkenbein , a. i. larkin , and v. m. vinokur , rev .
physics * 66 * , 1125 ( 1994 ) . v.
m. vinokur , m. v. feigelman , v. b. geshkenbein , phys .
lett . * 67 * , 915 ( 1991 ) .
e. zeldov , n. m. amer , g. koren , a. gupta , m. w. mcelfresh , and r. j. gambino , appl .
lett . * 56 * , 680 ( 1990 ) .
m. p. maley , j. o. willis , h. lessure , m. e. mchenry , phys .
b * 42 * , 2639 ( 1990 ) . m. e. mchenry , s. simizu , h. lessure , m. p. maley , j. y. coulter , i. tanaka , h. kojima , phys . rev .
b * 44 * , 7614 ( 1991 ) .
lu zhang , j. z. liu , r. n. shelton , m. d. lan , phys . rev .
b * 50 * , 7092 ( 1994 ) .
r. viennois , e. giannini , d. van der marel , r. cerny , journal of solid state chemistry * 183 * , 769 - 775 ( 2010 ) .
t. j. liu , x. ke , b. qian , j. hu , d. fobes , e. k. vehstedt , h. pham , j. h. yang , m. h. fang , l. spinu , p. schiffer , y. liu , z. q. mao , phys .
b * 80 * , 174509 ( 2009 ) .
h. okamoto , l. e. tanner , bull . alloy phase diagrams * 11 * ( 1990 ) .
e. c. stoner , philosophical magazine * 36 * , 803 ( 1945 ) .
a. k. pramanik , l. harnagea , c. nacke , a. u. b. wolter , s. wurmehl , v. kataev , b. bchner , phys .
b * 83 * , 094502 ( 2011 ) .
a. a. zhukov , h. kpfer , g. perkins , l. f. cohen , a. d. caplin , s. a. klestov , h. claus , v. i. voronkova , t. wolf , h. whl , phys .
b * 51 * , 12704 ( 1995 ) . c. p. bean , phys . rev .
* 8 * , 250 ( 1962 ) . c. p. bean , rev .
phys . * 36 * , 31 ( 1964 ) .
t. taen , y. tsuchiya , y. nakajima , t. tamegai , phys .
b * 80 * , 092502 ( 2009 ) .
y. liu , r. k. kremer , c. t. lin , eur .
. lett . * 92 * , 57004 ( 2010 ) .
r. griessen , wen hai - hu , a. j. j. van dalen , b. dam , j. rector , h. g. schnack , s. libbrecht , e. osquiguil , y. bruynseraede , phys .
lett . * 72 * , 1910 ( 1994 ) .
w. a. fietz , w. w. webb , phys . rev . * 178 * , 657 ( 1969 ) .
d. dew - hughes , phil . mag . *
30 * , 293 ( 1974 ) .
m. r. koblischka , a. j. j. van dalen , t. higuchi , s. i. yoo , m. murakami , phys .
b * 58 * , 2863 ( 1998 )
. e. j. kramer , j. appl .
phys . * 44 * , 1360(r ) ( 1973 ) .
j. w. ekin , _ experimental techniques for low - temperature measurements _ , oxford university press ( new york ) 2006 pp 412 - 467
. j. w. ekin , cryogenics * 20 * , 611 ( 1980 ) .
r. flkiger , c. senatore , m. cesaretti , f. buta , d. uglietti , b. seeber , supercond .
* 21 * , 054015 ( 2008 ) .
m. putti , i. pallecchi , e. bellingeri , m. r. cimberle , m. tropeano , c. ferdeghini , a. palenzona , c. tarantini , a. yamamoto , j. jiang , j. jaroszynski , f. kametani , d. abraimov , a. polyanskii , j. d. weiss , e. e. hellstrom , a. gurevich , d. c. larbalestier , r. jin , b. c. sales , a. s. sefat , m. a. mcguire , d. mandrus , p. cheng , y. jia , h. h. wen , s. lee and c. b. eom , supercond .
* 23 * , 034003 ( 2010 ) .
d. b. montgomery , w. sampson , appl .
* 6 * , 108 ( 1965 ) .
y. shapira , l. j. neuringer , phys .
rev . a * 140 * , 163 ( 1965 ) .
m. r. koblischka , m. murakami , supercond .
* 13 * , 738 - 744 ( 2000 ) . c. tarantini , s. lee , y. zhang , j. jiang1 , c. w. bark , j. d. weiss , a. polyanskii , c. t. nelson , h. w. jang , c. m. folkman , s. h. baek , x. q. pan , a. gurevich , e. e. hellstrom , c. b. eom , and d. c. larbalestier , appl .
. lett . * 96 * , 142510 ( 2010 ) .
hefei hu , jian - min zuo , jinsheng wen , zhijun xu , zhiwei lin , qiang li , genda gu , wan kyu park , laura h greene , new journal of physics * 13 * , 053031 ( 2011 ) .
xiaobo he , guorong li , jiandi zhang , a. b. karki , rongying jin , b. c. sales , a. s. sefat , m. a. mcguire , d. mandrus , e. w. plummer , phys .
b * 83 * , 220502(r ) ( 2011 ) .
r. khasanov , m. bendele , a. amato , p. babkevich , a. t. boothroyd , a. cervellino , k. conder , s. n. gvasaliya , h. keller , h .- h .
klauss , h. luetkens , v. pomjakushin , e. pomjakushina , b. roessli , phys .
b * 80 * , 140511(r ) ( 2009 ) .
l. li , z. r. yang , z. t. zhang , l. pi , s. tan , y. h. zhang , new journal of physics * 12 * , 063019 ( 2010 ) . n. katayama , s. ji , d. louca , s. lee , m. fujita , t. j. sato , j. wen , z. xu , g. gu , g. xu , z. lin , m. enoki , s. chang , k. yamada , j. m. tranquada , j. phys .
jpn . * 79 * , 113702 ( 2010 ) .
m. bendele , p. babkevich , s. katrych , s. n. gvasaliya , e. pomjakushina , k. conder , b. roessli , a. t. boothroyd , r. khasanov , h. keller , phys .
b * 82 * , 212504 ( 2010 ) . c. s. yadav , p. l. paulose , journal of physics : conference series * 273 * , 012108 ( 2011 ) .
huan yang , cong ren , lei shan , and hai - hu wen , phys .
b * 78 * , 092504 ( 2008 ) . h. g. schnack , r. griessen , j. g. lensink , c. j. van der beek , p. h. kes , physica c * 197 * , 337 - 361 ( 1992 ) .
m. jirsa , l. pust , h. g. schnack , r. griessen , physica c * 207 * , 85 - 96 ( 1993 ) . | an extensive study of the magnetic properties of fete@xmath0se@xmath1 crystals in the superconducting state is presented .
we show that weak collective pinning , originating from spatial variations of the charge carrier mean free path ( @xmath2 pinning ) , rules in this superconductor .
our results are compatible with the nanoscale phase separation observed on this compound and indicate that in spite of the chemical inhomogeneity spatial fluctuations of the critical temperature are not important for pinning .
a power law dependence of the magnetization vs time , generally interpreted as signature of single vortex creep regime , is observed in magnetic fields up to 8 t. for magnetic fields applied along the @xmath3 axis of the crystal the magnetization curves exhibit a clear peak effect whose position shifts when varying the temperature , following the same dependence as observed in yba@xmath4cu@xmath5o@xmath6 .
the time and temperature dependence of the peak position has been investigated .
we observe that the occurrence of the peak at a given magnetic field determines a specific vortex configuration that is independent on the temperature .
this result indicates that the influence of the temperature on the vortex - vortex and vortex - defect interactions leading to the peak effect in fete@xmath0se@xmath1 is negligible in the explored range of temperatures . |
lorentz invariance is one of the foundations of the standard model of particle physics .
the constraints on possible lorentz violating phenomenology are quite severe , see for example , the summary tables that provided by kostelecky _
et al_.@xcite .
the gravitational interaction is far more weak , compare to other fundamental interactions .
this allows one to study the possible lorentz violating effects on certain gravity theories , such as einstein - aether theory @xcite and horava - lifshitz theory @xcite .
one feature of lorentz invariance violation is that the speed of light differ from the one in special relativity .
the gravity theories with lorentz violation could have the feature that the speed of graviton or the speed of gravitational wave differ from the one in general relativity . studying
the speed of gravitational wave in a lorentz violating gravity theory will give different perspective on quantum gravitational phenomena .
one of the most important prediction of einstein s general relativity is gravitational radiation .
many pioneer works @xcite have discussed the gravitational radiation in both theoretical properties and experimental approaches of detections . currently , the most sensitive measurement is provided by ground - based laser interferometer gravitational - wave observatory ( ligo ) detector @xcite .
another sensitive measurement , which is in progress , is the laser interferometer space antenna ( lisa ) that detect and accurately measure gravitational waves from astronomical sources .
the primordial gravitational waves @xcite could be of interest to cosmologists as they provide a new and unique window on the earliest moments in the history of the universe and on possible new physics at energies many orders of magnitude beyond those accessible at particle accelerators . in general relativity ,
the effects of gravitation are ascribed to spacetime curvature instead of a force . however , up to now , general relativity still faces problems .
first , the recent astronomical observations @xcite found that our universe is accelerated expanding . this result can not be obtained directly from einstein s gravity and his cosmological principle . since normal matters only provide attractive force .
the most widely adopted way to resolve it is involving the so called dark energy which provides the repulsive force .
second , the flat rotation curves of spiral galaxies violate the prediction of einstein s gravity @xcite .
the most widely adopted way to resolve it is involving the so called dark matter which provides enough attractive force such that the discrepancy is restored .
the above astronomical phenomena occur at very large cosmological scale .
the following anomalies occur in solar system which imply the newton s inverse - square law of universal gravitation and general relativity need modifications .
the third one , two pioneer spacecrafts suffer an anomalous constant sunward acceleration , @xmath0 @xcite .
the fourth one , it has been observed at various occasions that satellites after an earth swing - by possess a significant unexplained velocity increase by a few mm / s @xcite . the fifth one , from the analysis of radiometric measurements of distances between the earth and the major planets including observations from martian orbiters and landers from 1961 to 2003 a secular trend of the astronomical unit of @xmath1 m / cy has been reported @xcite .
the sixth one , a recent orbital analysis of lunar laser ranging ( llr ) @xcite shows an anomalous secular eccentricity variation of the moon s orbit @xmath2 .
all the facts imply that the einstein s theory should be modified . by mimicking einstein ,
we have proposed that the modified gravitational theory should correspond to a new geometry which involves riemann geometry as its special case .
finsler geometry @xcite as a nature extension of riemann geometry is a good candidate to solve the problems mentioned above .
a new geometry ( finsler geometry ) involves new spacetime symmetry .
the lorentz violation is intimately linked to finsler geometry .
kostelecky @xcite have studied effective field theories with explicit lorentz violation in finsler spacetime .
finsler geometry really gives better description for the nature of gravity : the flat rotation curves of spiral galaxies can be deduced naturally without invoking dark matter @xcite ; a finlerian gravity model could account for the accelerated expanding university without invoking dark energy @xcite ; a special finsler space - randers space @xcite could account for the anomalous acceleration @xcite in solar system observed by pioneer 10 and 11 spacecrafts ; the finsler spacetime could give a modification on the gravitational deflection of light @xcite , which may account to these observations without adding dark matter in bullet cluster @xcite ; the result based on the kinematics with a special finsler spacetime is in good agreement with secular trend of the astronomical unit and secular eccentricity variation of the moon s orbit @xcite .
it is interest to investigate the gravitational wave in finsler spacetime .
it is well known that the gravitational wave propagates with the speed of light in general relativity .
this is due to the fact that the spacetime metric is close to the minkowski metric in the weak gravitational field approximation , and the causal speed of minkowski spacetime is just the speed of light . however , in finsler spacetime the causal speed is generally different with the speed of light @xcite . in this paper
, we will present the solution of linearized gravitational vacuum field equation in finsler spacetime .
it is shown that there is possibility that the causal speed of it is larger than the speed of light .
instead of defining an inner product structure over the tangent bundle in riemann geometry , finsler geometry is based on the so called finsler structure @xmath3 with the property @xmath4 for all @xmath5 , where @xmath6 represents position and @xmath7 represents velocity . the finsler metric is given as @xcite @xmath8 finsler geometry has its genesis in integrals of the form @xmath9 the finsler structure represents the length element of finsler space .
the parallel transport has been studied in the framework of cartan connection @xcite .
the notation of parallel transport in finsler manifold means that the length @xmath10 is constant . the geodesic equation for finsler manifold is given as @xcite @xmath11 where @xmath12 is called geodesic spray coefficient
. obviously , if @xmath3 is riemannian metric , then @xmath13 where @xmath14 is the riemannian christoffel symbol .
since the geodesic equation ( [ geodesic ] ) is directly derived from the integral length @xmath15 the inner product @xmath16 of two parallel transported vectors is preserved . in finsler manifold , there exists a linear connection - the chern connection @xcite .
it is torsion freeness and almost metric - compatibility , @xmath17 where @xmath18 is the formal christoffel symbols of the second kind with the same form of riemannian connection , @xmath19 is defined as @xmath20 and @xmath21 is the cartan tensor ( regarded as a measurement of deviation from the riemannian manifold ) . in terms of chern connection , the curvature of finsler space is given as @xmath22 where @xmath23 .
the gravity in finsler spacetime has been investigated for a long time @xcite . in this paper , we introduce vacuum field equation by the way discussed first by pirani @xcite . in newton s theory of gravity , the equation of motion of a test particle is given as @xmath24 where @xmath25 is the gravitational potential and @xmath26 is euclidean metric .
for an infinitesimal transformation @xmath27(@xmath28 ) , the equation ( [ dynamic newton ] ) becomes , up to first order in @xmath29 , @xmath30 combining the above equations([dynamic newton ] ) and ( [ dynamic newton1 ] ) , we obtain @xmath31 in newton s theory of gravity , the vacuum field equation is given as @xmath32 .
it means that the tensor @xmath33 is traceless in newton s vacuum . in general relativity ,
the geodesic deviation gives similar equation @xmath34 where @xmath35 .
here , @xmath36 is riemannian curvature tensor , @xmath37 denotes the covariant derivative alone the curve @xmath38 .
the vacuum field equation in general relativity gives @xmath39@xcite .
it implies that the tensor @xmath40 is also traceless , @xmath41 . in finsler spacetime
, the geodesic deviation gives @xcite @xmath42 where @xmath43 .
here , @xmath44 is finsler curvature tensor defined in ( [ finsler curvature ] ) , @xmath37 denotes covariant derivative @xmath45 . since the vacuum field equations of newton s gravity and general relativity have similar form , we may assume that vacuum field equation in finsler spacetime hold similar requirement as the case of netwon s gravity and general relativity .
it implies that the tensor @xmath46 in finsler geodesic deviation equation should be traceless , @xmath47 .
we have proved that the analogy from the geodesic deviation equation is valid at least in finsler spacetime of berwald type @xcite .
for this reason , we may suppose that this analogy is valid in general finsler spacetime .
it should be noticed that @xmath48 is called the ricci scaler , which is a geometrical invariant . for a tangent plane @xmath49 and a non - zero vector @xmath50 ,
the flag curvature is defined as @xmath51 where @xmath52 .
the flag curvature is a geometrical invariant that generalizes the sectional curvature of riemannian geometry .
it is clear that the ricci scaler @xmath53 is the trace of @xmath46 , which is the predecessor of flag curvature .
therefore , the value of ricci scaler @xmath53 is invariant under the ordinate transformation .
furthermore , the predecessor of flag curvature could be written in terms of the geodesic spray coefficient @xmath54 thus , the ricci scaler @xmath53 is insensitive to connection that one is using , it only depends on the length element @xmath3 .
the gravitational vacuum field equation @xmath55 is universal in any types of theories of finsler gravity .
et al_. @xcite have constructed gravitational dynamics for finsler spacetimes in terms of an action integral on the unit tangent bundle .
their researches also show that the gravitational vacuum field equation in finsler spacetime is @xmath55 .
it is hard to find a non trivial solution of the gravitational vacuum field equation ( @xmath55 ) in finsler spacetime . here , we study the weak field radiative solution of the finslerian vacuum field equation @xmath55 .
it is well known that the minkowski spacetime is a trivial solution of einstein s vacuum field equation . in the finsler spacetime
, the trivial solution of finslerian vacuum field equation is called locally minkowski spacetime .
a finsler spacetime is called a locally minkowshi spacetime if there is a local coordinate system @xmath56 , with induced tangent space coordinates @xmath57 , such that @xmath3 depends only on @xmath58 and not on @xmath6 . using the formula ( [ predecessor flag curvature ] ) , one knows obvious that locally minkowski spacetime is a solution of finslerian vacuum field equation .
we suppose that the metric is close to the locally minkowski metric @xmath59 , @xmath60 where @xmath61 . in the rest of this section ,
the lowering and raising of indices are carried out by @xmath62 and its matrix inverse @xmath63 .
to first order in @xmath64 , the geodesic spray coefficient is @xmath65 we have already used the euler s theorem for homogeneous functions to obtain the above equation . and the ricci scaler is @xmath66 where @xmath67 and @xmath68 since the value of ricci scaler @xmath53 is invariant under the coordinate transformation , we must fix the gauge symmetry to yield unique solution . under a coordinate transformation
@xmath69 the metric @xmath70 transforms as @xmath71 by performing the coordinate transformation with @xmath72 we find that @xmath73 satisfies @xmath74 this choice of gauge ( [ gauge ] ) has the same form with the lorentz gauge in general relativity , due to the fact that the locally minkowshi metric @xmath62 does not depend on @xmath6 . by making use of the finslerian gauge ( [ gauge ] ) , and noticing that @xmath62 does not depend on @xmath6 ,
we rewrite the ricci scaler as @xmath75 we find from ( [ ricci symplify ] ) that the solution of @xmath55 has following properties @xmath76 where @xmath77 @xmath78 is function of @xmath58 and @xmath79 is the polarization tensor .
the term @xmath80 could be written as @xmath81 where @xmath82 . substituting the equation ( [ det a ] ) into ( [ constrain k ] ) , we obtain @xmath83
it is obvious that @xmath84 while the finsler spacetime @xmath62 is not minkowskian .
it implies that the wave vectors @xmath85 of gravitational waves is not null in finsler spacetime @xmath62 .
the randers spacetime @xcite is a special kind of finsler geometry with finsler structure @xmath86 where @xmath87 and @xmath88 is riemannian metric .
the indices on certain objects that decorated with a bar are lowered and raised by @xmath88 and its matrix inverse @xmath89 . substituting the randers - finsler structure @xmath90 into the dispersion relation of gravitational wave ( [ dispersion relation ] ) and supposing
the randers spacetime is very close to minkowski spacetime @xmath88 , to first order in @xmath91 , we obtain @xmath92 where @xmath93 denotes the inner product on minkowski spacetime @xmath88 and @xmath94 .
the causality should holds in finsler spacetime @xmath62 , thus @xmath95 while the signature of minkowski metric @xmath88 is of the form @xmath96 .
if @xmath97 , it means that the speed of gravitational wave is larger than speed of light .
it implies that the speed of gravitational wave could larger than speed of light and causality still holds .
the inequalities @xmath95 and @xmath97 satisfy if @xmath98 so that the speed of gravitational wave in the anisotropic spacetime is larger than the speed of light and the causality still holds .
the sketch figure of the causal structure of finsler spacetime ( @xmath62 ) is shown in fig.1 .
it is clear from fig.1 that the null vectors on finsler spacetime ( @xmath62 ) are spacelike vectors on minkowski spacetime .
the causal speed of finsler spacetime could be larger than the one of minkowski spacetime . ) .
the dot line denotes the null structure on minkowski spacetime .
it is obvious that the null structure of finsler spacetime is larger than the one in minkowski spacetime ]
in this paper , we used the weak gravitational field approximation to get a linearized solution of the gravitational vacuum field equation in finsler spacetime .
the plane - waves solution ( [ h munu ] ) of gravitational wave in an anisotropic spacetime was presented .
it is shown that the gravitational wave is propagating in locally minkowski spacetime ( @xmath62 ) .
the killing vectors of locally minkowski spacetime ( @xmath62 ) are investigated in ref.@xcite .
it was shown that finsler spacetime admits less symmetry than minkowski spacetime , and the translation symmetry in preserved in locally minkowski spacetime ( @xmath62 ) .
based on the noether theorem , the spacetime translational invariance implies that the energy - momentum is well defined and conserved in locally minkowski spacetime ( @xmath62 ) .
the dispersion relation of gravitational wave in finsler spacetime ( [ dispersion relation ] ) was presented .
the speed of gravitational wave could larger than the speed of light in randers spacetime and the casuality of gravitational wave still holds , if the condition ( [ requirement ] ) is satisfied . since the wave vector @xmath85 of gravitational wave is timelike in locally minkowski spacetime ( @xmath62 ) , it would not lose energy via the gravitational cherenkov radiation .
we would like to thank m. h. li and s. wang for useful discussions .
one of us x. li thanks prof . c. pfeifer for usefol discussions .
the work was supported by the nsf of china under grant no .
10875129 and 11075166 and 11147176 .
999 v. a. kostelecky and n. russell , arxiv:0801.0287[hep - ph ] .
t. jacobson and d. mattingly , phys .
d * 64 * , 024028 ( 2001 ) .
p. hoava , phys .
d * 79 * , 084008 ( 2009 ) ; d. vernieri and t. p. sotiriou , arxiv:1112.3385[hep - th ] .
v. b. braginsky and v. n. rudenko , physics reports * 46 * , 165 ( 1978 ) .
k. s. thorne , rev .
phys . * 52 * , 299 ( 1980 ) .
r. weiss , rev .
phys . * 71 * , s187 ( 1999 ) .
b. p. abbott _
arxiv:0909.3583v4 [ astro-ph.he ] . l. m. krauss , s. dodelson and s. meyer , science * 328 * , 989 ( 2010 ) .
l. p. grishchuk , sov .
jetp lett .
* 23 * , 293 ( 1976 ) .
a. g. riess , r. p. kirshner , b. p. schmidt , et al .
astrophys j. * 117 * , 707 ( 1999 ) ; s. perlmutter , g. aldering , g. goldhaber , et al .
astrophys j. * 517 * , 565 ( 1999 ) ; c. l. bennett , m. halpern , g. hinshaw , et al .
astrophys j. * 148 * ( suppl ) , 1 ( 2003 ) .
f. zwicky , helv .
acta * 6 * , 110 ( 1933 ) .
j. d. anderson , _ et al_. , phys
lett . * 81 * , 2858 ( 1998 ) ; phys .
d * 65 * , 082004 ( 2002 ) ; mod .
a * 17 * , 875 ( 2002 ) .
j. d. anderson , _ et al_. , phys .
100 * , 091102 ( 2008 ) .
g. a. krasinsky and v. a. brumberg , celest .
. astrn . * 90 * , 267 ( 2004 ) .
j. g. williams and d. h. boggs , _ in proceedings of 16th international workshop on laser ranging _ , ed .
s. schillak , ( space research centre , polish academy of sciences ) , 2009 .
d. bao , s. s. chern and z. shen , _ an introduction to riemann finsler geometry _ , graduate texts in mathematics * 200 * , springer , new york , 2000 .
v. a. kostelecky , phys .
d * 69 * , 105009 ( 2004 ) ; phys .
b * 701 * , 137 ( 2011 ) .
z. chang and x. li , phys .
b * 668 * , 453 ( 2008 ) ; x. li and z. chang , arxiv:1108.3443v1 [ gr - qc ] .
z. chang and x. li , phys .
b * 676 * , 173 ( 2009 ) ; x. li , z. chang and m. h. li , arxiv:1001.0066v1 [ gr - qc ] ; z. chang , m. h. li and x. li , arxiv:1009.1509v1 [ gr - qc ] , x. li and z. chang , phys .
b * 692 * , 1 ( 2010 ) .
x. li and z. chang , phys .
d * 82 * , 124009 ( 2010 ) .
d. clowe , s. w. randall , and m. markevitch , http:// flamingos.astro.ufl.edu/1e0657/index.html ; nucl .
suppl . * 173 * , 28 ( 2007 ) .
x. li and z. chang , cpc ( hep & np ) * 35 * , 914 ( 2011 ) ; arxiv:0911.1890v1 [ gr - qc ] .
m. matsumoto , _ foundations of finsler geometry and special finsler spaces _ , kaiseisha press , saikawa shigaken , japan 1986 .
p. l. antonelli and s. f. rutz , _ finsler geometry _
, advanced studies in pure mathematics 48 , sapporo ( 2005 ) p. 210
-in memory of m.matsumoto .
z. szabo , ann .
geom * 34 * , 381 ( 2008 ) .
s. s. chern , sci .
tsing hua univ .
a * 5 * , 95 ( 1948 ) ; or selected papers , vol .
ii , 194 , springer 1989 .
y. takano , lett .
nuovo cimento , * 10 * , 747 ( 1974 ) .
s. ikeda , ann .
, * 44 * , 558 ( 1987 ) .
r. tavakol , n. van den bergh , phys .
a * 112 * , 23 ( 1985 ) .
g. yu . bogoslovsky , phys .
, * 24 * , 354 ( 1993 ) .
f. a. e. pirani , lectures on general relativity , brandeis summer institute in theoretical physics , vol . 1 , 1964 . s. f. rutz , computer physcis communications * 115 * , 300 ( 1998 ) .
s. weinberg , _ gravitation and cosmology : principles and applications of the general theory of relativity _ , john wiley & sons , new york , 1972 .
c. pfeifer and m. n.r .
wohlfarth , arxiv:1112.5641[gr - qc ] .
g. randers , phys . rev .
* 59 * , 195 ( 1941 ) . | by making use of the weak gravitational field approximation , we obtain a linearized solution of the gravitational vacuum field equation in an anisotropic spacetime .
the plane - wave solution and dispersion relation of gravitational wave is presented explicitly .
there is possibility that the speed of gravitational wave is larger than the speed of light and the casuality still holds .
we show that the energy - momentum of gravitational wave in the ansiotropic spacetime is still well defined and conserved . |
this paper presents a unified mathematical framework for probabilistic inference with statistical models , such as graphical models .
our approach is summarized as follows : * ( a ) statistical models are algebraic varieties .
* * ( b ) every algebraic variety can be tropicalized . * * ( c ) tropicalized statistical models are fundamental for parametric inference . * by a _ statistical model _ we mean a family of joint probability distributions for a collection of discrete random variables @xmath0 .
thesis ( a ) states that many families of interest can be characterized by polynomials in the joint probabilities @xmath1 .
the emerging field of algebraic statistics @xcite offers algorithms for this polynomial representation .
_ tropicalization _ means replacing the arithmetic operations @xmath2 by the operations @xmath3 .
this process captures the essence of what happens when the joint probabilities @xmath4 are replaced by their logarithms .
the tropicalization of an algebraic variety is a piecewise - linear set which enjoys many features familiar from algebraic geometry @xcite . in particular ,
the tropicalization of a statistical model is a piecewise - linear set in the space with logarithmic coordinates @xmath5 .
thesis ( c ) states that tropical algebraic geometry of statistical models is fundamental in analyzing the behavior of inference algorithms under the variation of model parameters . by _ inference
_ we mean the evaluation of one or more coordinates of a single point on the algebraic variety , in either @xmath2 or @xmath3 arithmetic .
this is the standard notion of inference used for graphical models in statistical learning theory @xcite , but it differs from other ( more classical ) notions of inference in mathematical statistics . by _ parametric inference _ we mean the analysis of the dependence of inference on parameters . to give a more concrete discussion of parametric inference it is useful to focus
on directed graphical models .
a _ directed graphical model _ ( or _ bayesian network _ ) is a finite directed acyclic graph @xmath6 with two kinds of vertices , _ observed variables _ @xmath0 and _ hidden variables _
@xmath7 , where each edge is labeled by a transition matrix whose entries are linear forms in some parameters .
the rules of discrete probability express the observed probabilities @xmath8 as polynomials of degree @xmath9 in the parameters , where @xmath9 is the number of edges of @xmath6 .
the polynomials parametrize the graphical model as an algebraic variety .
the two standard types of inference questions for graphical models are : 1 . the calculation of _ marginal probabilities _ : @xmath10 2 .
the calculation of _ maximum a posteriori ( map ) _ log probabilities : @xmath11 where the @xmath12 range over all the possible assignments for the hidden random variables @xmath13 .
together , these two primitives can be used to effectively solve a range of other inference problems , including the calculation of conditional probabilities and other quantities of interest .
the key to inference in graphical models is the _ sum - product algorithm _
@xcite ( also known as the _ generalized distributive law _
this polynomial - time algorithm is used , both in ordinary arithmetic @xmath2 and in tropical arithmetic @xmath3 , to _ efficiently _ solve problems 1 and 2 . for more background on the sum - product algorithm , and for connections to message passing and the junction tree algorithm
see @xcite .
although the sum - product algorithm provides efficient solutions to the basic inference problems 1 and 2 , it only applies to one coordinate @xmath8 of one distribution at a time .
what we are interested in are the _ parametric _ versions of the inference problems .
they can be phrased as follows : 1 .
find all parameters for a model which result in the same values for all @xmath14 .
2 . given observations @xmath15 and hidden data @xmath16 , identify all parameters such that @xmath17 is the most likely explanation for the observations @xmath18 . as we will see , the following _ modeling _ questions are fundamentally related to problems 3 and 4 : 1 .
which ( parameter independent ) relations on the probabilities @xmath8 does the model imply ? 2 .
describe the tropicalization of the variety corresponding to a graphical model . problem 5 asks for the ideal of _ polynomial invariants _ of a statistical model @xcite .
invariants have been investigated in phylogenetics @xcite where they can help to identify good trees for aligned dna sequences .
the primary goal of our study is to give a practical answer to question 4 for graphical models .
our main algorithmic result is an efficient procedure for parametric inference that can be viewed as a polytopal analog of the sum - product algorithm .
the efficiency is based on the complexity estimates for newton polytopes which we derive in section 4 .
the resulting _ polytope propagation algorithm _ is applied to problems in biological sequence analysis in the companion paper @xcite .
the mathematics to be developed in sections 3 and 4 is of independent interest .
it also furnishes new tools for parametric inference ( problems 3 and 4 ) and parametric modeling ( problems 5 and 6 ) which are applicable to a wide range of statistical problems .
we demonstrate this by analyzing the hidden markov model ( hmm ) and the general markov model on a binary tree , in sections 2 and 5 respectively .
a graphical model is an algebraic variety which is presented as the image of a highly structured polynomial map @xmath19 . here
@xmath20 is the space whose coordinates are the model parameters @xmath21 and @xmath22 is the space whose coordinates @xmath23 are the joint probabilities for the observed random variables . in applications ,
the integer @xmath24 is much larger than the integer @xmath25 , in fact ; it is so large that one can only look at one coordinate @xmath26 at a time .
each coordinate @xmath27 of the map @xmath28 is a polynomial function in @xmath21 .
the efficient evaluation of these functions relies on the sum - product algorithm .
here we study the ( parametric ) inference and modeling problems in the familiar context of the _ hidden markov model _ ( hmm ) .
a discrete hmm has @xmath29 observed states @xmath30 taking on @xmath31 possible values , and @xmath29 hidden states @xmath32 taking on @xmath33 possible values .
the hmm can be characterized by the following conditional independence statements for @xmath34 : @xmath35 we consider the homogeneous model with uniform initial distribution , where all transitions @xmath36 are given by the same @xmath37-matrix @xmath38 and all transitions @xmath39 are given by the same @xmath40-matrix @xmath41 . throughout our discussion we disregard for simplicity the usual probabilistic hypothesis that @xmath42 and @xmath43 are non - negative and all row sums are @xmath44 . [ hmmprop ] the hidden markov model is the image of a map @xmath45 , where @xmath46 and each coordinate of @xmath28 is a bi - homogeneous polynomial of degree @xmath47 in @xmath42 and degree @xmath29 in @xmath43 .
problem 3 is to compute the fibers of the map @xmath28 . in statistics ,
this is called _
parameter identification_. we use the term _ coordinate polynomials _ for the polynomials @xmath48 that are coordinates of the map @xmath28 .
our running example in this section is the case @xmath49 with binary random variables @xmath50
. the graph of this model is drawn in figure [ fig : hmm ] .
the shaded nodes are the observed random variables . here
the parameter space is @xmath51 with coordinates @xmath52 , and it maps to @xmath51 with coordinates @xmath53 .
the map @xmath54 is given by @xmath55 the hidden markov model ( i.e. the image of @xmath28 ) is the zero set of the quartic polynomial @xmath56 this polynomial was found by a _
grbner basis _ computation .
see the discussion on _ implicitization _ in @xcite
. in general , the polynomial functions on @xmath57 which vanish on the image of @xmath28 are the called _ invariants of the model_. they form a prime ideal @xmath58 . in our example
, @xmath58 is generated by the quartic polynomial above .
problem 5 is to compute generators of the ideal @xmath58 . when @xmath59 and @xmath25 are small , this can be done using grbner bases , and in some cases it is possible to characterize @xmath58 based on the structure of the model ( see , for example , conjecture [ luckydeterminants ] ) , but in general problem 5 is hard and the ideal @xmath58 may remain unknown .
here is where tropical geometry comes in .
the _ tropicalization _ of our map @xmath28 is the map @xmath60 defined by replacing products by sums and sums by minima in the formula for @xmath28 . in our example @xmath61
, the tropicalization is the piecewise - linear map @xmath62 with @xmath63 this minimum is attained by the most likely hidden data @xmath64 , given the observations @xmath65 and given the parameters @xmath66 and @xmath67 . the sequence @xmath64 is known as the _ viterbi sequence _ in the hmm literature @xcite .
it solves problem 2 in the introduction .
the key observation , which we discuss in more detail in section 4 , is that the set of parameters @xmath68 which select the viterbi sequence @xmath69 is the normal cone at a vertex of the newton polytope of the polynomial @xmath70 .
this polytope is @xmath71-dimensional , it has @xmath72 vertices , and its normal fan represents the solution to problem 4 in the introduction when @xmath73 is fixed .
we can also consider an extension of problem 4 where @xmath73 ranges over all possible observations .
the solution is given by the newton polytope of the map @xmath28 .
in our example , this is a @xmath74-dimensional polytope with @xmath75 vertices , @xmath76 edges , @xmath77 two - faces , @xmath78 three - faces and @xmath79 facets , namely , the minkowski sum of eight copies of the earlier @xmath71-dimensional polytope for @xmath80 . for a concrete numerical example , fix the parameters @xmath81 and @xmath82 .
we find : @xmath83 the set of all parameters @xmath84 leading to the same conclusions as @xmath85 is the cone defined by @xmath86 our solution to the parametric inference problem with respect to all observations simultaneously consists of @xmath75 such cones .
the _ tropical hmm _ is the union of the images of these cones under the piecewise - linear map @xmath87 .
this image is a piecewise - linear set of dimension @xmath88 .
the cone which contains the chosen parameters @xmath89 mapped to a @xmath88-dimensional cone in the tropical hmm ( it spans the hyperplane @xmath90 ) but most of the other @xmath91 cones are mapped to lower - dimensional cones by the map @xmath92 .
the question how the number @xmath75 grows as the length @xmath29 increases will be addressed in corollary 10 .
we have seen that a graphical model is the image of a polynomial map @xmath28 from the space of parameters to the space of joint probability distributions on the observed random variables .
furthermore , we have seen that the tropicalization of @xmath28 arises naturally in solving problem 4 . in this section
we study the geometry of tropicalization in the more general setting where @xmath93 is an arbitrary polynomial map . in statistical applications , it is usually the case that each coordinate @xmath48 of the map @xmath28 is a polynomial with positive coefficients . if this holds then the polynomial map @xmath28 is called _
positive_. we say that @xmath28 is _ surjectively positive _ if , in addition , @xmath28 maps the positive orthant surjectively onto the positive points in the image , in symbols , @xmath94 the set of all polynomial functions which vanish on the image of @xmath28 is a prime ideal @xmath58 in the polynomial ring @xmath95 $ ] .
the closure of the image of @xmath28 is the variety of the prime ideal @xmath58 . in tropical geometry
, we replace the variety of @xmath58 by a piecewise - linear set as follows .
the _ tropical variety _
@xmath96 is the set of all weight vectors @xmath97 such that the initial ideal @xmath98 contains no monomial @xcite .
following @xcite , we define the _ positive tropical variety _ @xmath99 as the set of all weight vectors @xmath97 such that the initial ideal @xmath98 contains no polynomial with only positive coefficients .
the tropical variety @xmath96 is a _
polyhedral fan _ in @xmath22 , and @xmath99 is a _
polyhedral subcomplex _ of @xmath96 .
this means that @xmath96 is a finite union of closed convex polyhedral cones that fit together nicely , and @xmath99 is the union of a subset of these cones .
the _ tropicalization _ of the polynomial map @xmath28 is the piecewise - linear map @xmath100 defined by replacing products by sums and sums by minima in the evaluation of @xmath28 .
we say that @xmath92 is a _ tropical morphism_. examples of tropical morphisms appear in the displayed formulas ( [ maxformula ] ) , ( [ formula1 ] ) , ( [ formula2 ] ) , ( [ formula3 ] ) , ( [ tropicaltreemap ] ) and ( [ treemapexample3 ] ) .
the following theorem describes the geometry of this situation .
we define the _ newton polytope _ of a polynomial map @xmath101 as the minkowski sum in @xmath102 of the newton polytopes of its coordinates @xmath103 . for basics on newton polytopes and
their normal fans see @xcite .
[ hmmmain ] the tropical morphism @xmath92 is linear on each cone in the normal fan of the newton polytope of @xmath28 .
its image is a fan contained in @xmath96 .
if @xmath28 is positive then @xmath104 is a subset of @xmath99 , but it is generally not a polyhedral subcomplex . if @xmath28 is surjectively positive then @xmath105 .
let @xmath106 denote the newton polytope of the polynomial @xmath107 .
by definition , @xmath106 is the convex hull in @xmath102 of all non - negative lattice points @xmath108 such that the monomial @xmath109 appears with non - zero coefficient in @xmath110 .
the piecewise - linear concave function @xmath111 is the _ support function _ of the polytope @xmath106 .
this means that @xmath112 is the minimum value attained on @xmath106 by the linear functional @xmath113 .
in particular , the function @xmath114 is linear on each cone in the normal fan of @xmath106 .
the newton polytope of the map @xmath28 is the minkowski sum @xmath115 .
the normal fan of @xmath116 is the common refinement of the normal fans of @xmath117 .
this shows that the function @xmath118 is linear on each cone of the normal fan of the newton polytope of @xmath28 .
since @xmath92 is continuous , the image of @xmath92 is a closed polyhedral fan in @xmath22 . consider any vector @xmath119 .
we must show that @xmath120 lies in @xmath96 , and if @xmath28 is positive then @xmath120 lies in @xmath99 .
let @xmath121 be any polynomial in the ideal @xmath58 .
if we substitute @xmath122 into @xmath123 then we get zero . consequently
, if we substitute the initial forms @xmath124 into the initial form @xmath125 then the result is zero .
see equation ( 11.2 ) on page 100 in @xcite .
this implies that @xmath125 is not a monomial .
moreover , if @xmath28 is positive then @xmath121 must have two terms whose coefficients have opposite signs .
this implies the desired conclusion .
the following example shows that @xmath126 need not be a subcomplex of @xmath99 .
if @xmath28 is assumed to be surjectively positive , then it follows from ( * ? ? ?
* proposition 2.5 ) that @xmath127 .
let @xmath128 , @xmath129 and consider the linear map @xmath130 then @xmath131 is the principal ideal generated by the linear form @xmath132 , and @xmath96 is essentially the normal fan of a tetrahedron .
we identify @xmath96 with the complete graph @xmath133 .
the six edges of @xmath133 are labeled with six monomial - free initial ideals of @xmath58 , namely , @xmath134 the first two of these six initial ideals contain a polynomial with positive coefficients . hence the positive tropical variety @xmath99 is the four - cycle in @xmath133 formed by the remaining four edges .
the tropicalization of the linear map @xmath28 is the tropical morphism @xmath135 the image of @xmath92 is the set of all vectors @xmath136 with @xmath137 .
each vector @xmath136 with @xmath138 has the initial ideal @xmath139 , so it lies on a particular edge of @xmath133 .
but the same edge also accounts for all vectors @xmath136 with @xmath140 ,
none of which is in the image of @xmath92 .
thus @xmath141 is a closed segment which covers only half of the edge of @xmath133 indexed by @xmath142 .
here it is easy to replace @xmath28 by a parameterization @xmath143 which is surjectively positive , for instance , @xmath144 @xmath145 we have @xmath146 but now the tropical morphism @xmath147 maps onto the entire four - cycle @xmath99 . in the rest of this section
we examine theorem [ hmmmain ] for a small but important graphical model , namely , the _ naive bayes model with two features _ @xcite .
there are two observed random variables @xmath148 and @xmath149 dependent on one hidden binary random variable @xmath150 .
the two observed variables take @xmath33 and @xmath31 possible values respectively .
the parameterization @xmath28 of this model is the map @xmath151 given by @xmath152 thus the model consists of all @xmath40-matrices @xmath153 of the form @xmath154 where @xmath155 is a @xmath156-matrix and @xmath157 is a @xmath158-matrix , i.e. , the model consists of precisely the @xmath40-matrices of rank @xmath159 .
[ threebythree ] the parameterization @xmath28 of the naive bayes model with two features is surjectively positive .
the ideal @xmath58 is generated by the @xmath160-subdeterminants of the @xmath40-matrix @xmath161 .
the map @xmath28 being positive means that if @xmath162 is any positive matrix of rank @xmath163 then @xmath42 and @xmath43 can be chosen to be positive .
this is a known result in linear algebra ( see e.g. @xcite ) .
the same statement is false for rank @xmath164 , i.e. , the parameterization of the naive bayes model with three or more features is not surjectively positive .
a well - known result in commutative algebra states that the @xmath165-minors of a @xmath40-matrix generate a prime ideal .
the variety of this ideal is the set of @xmath40-matrices of rank @xmath166 .
this our ideal @xmath58 for @xmath167 .
the objects of theorem [ hmmmain ] have been studied in @xcite and @xcite .
the tropical variety @xmath96 is the set of @xmath40-matrices of _ tropical rank _
@xmath159 , and the tropical variety @xmath168 is the set of @xmath40-matrices of _ barvinok rank _ @xmath159 .
develin @xcite determines the combinatorics and topology of these spaces when @xmath169 .
he shows that @xmath170 is shellable but @xmath171 can have torsion in its integral homology groups .
the newton polytope of the map @xmath28 is an interesting combinatorial object , namely , it is the @xmath172-dimensional zonotope associated with the complete bipartite graph @xmath173 .
the newton polytope of each coordinate @xmath174 is a line segment , and the zonotope is their minkowski sum .
the normal fan is the hyperplane arrangement @xmath175 .
its maximal cones correspond to the acyclic orientations of the complete bipartite graph @xmath173 .
west @xcite showed that the number of facets of such a cone can be any integer between @xmath176 and @xmath177 .
the total number of cones equals @xmath178 , where @xmath179 is the stirling number of the second kind . here
, the tropical morphism @xmath92 is given by @xmath180 the map @xmath181 is piecewise - linear with respect to the hyperplane arrangement .
recent work of federico ardilla ( in preparation ) gives a complete classification of all fibers of @xmath92 .
-0.3 cm -determinant . ]
-1.2 cm let @xmath182 , so the two observed random variables are ternary .
the prime ideal is @xmath183 the tropical variety @xmath96 is the fan over a two - dimensional polyhedral complex consisting of six triangles and nine quadrangles .
this complex is the @xmath163-skeleton of the product of two triangles , labeled as in figure 2a
. this complex is shellable .
the positive tropical variety @xmath99 is the subcomplex consisting of the nine quadrangles shown in figure 2b .
note that @xmath99 is a torus .
the newton polytope of @xmath28 is a five - dimensional zonotope with @xmath184 vertices , one for each acyclic orientation of the complete bipartite graph @xmath185 .
the map @xmath92 is linear on each of the @xmath184 cones in the corresponding hyperplane arrangement , but it is rank - deficient on @xmath186 of the cones .
the remaining @xmath187 cones are mapped onto the @xmath188 quadrangles of the torus @xmath99 .
thus the general fiber of @xmath92 involves @xmath189 cones .
of these , eight cones have @xmath74 facets , eight cones have @xmath190 facets , and two cones have @xmath188 facets .
consider a graphical model with @xmath9 edges and @xmath29 observed random variables @xmath30 each taking @xmath31 values .
such a model is given by a positive polynomial map @xmath191 .
each coordinate @xmath192 of @xmath28 is a polynomial of degree @xmath193 in the model parameters @xmath21 . in this section
we discuss the statistical meaning and the computational complexity of the mathematical objects introduced in the previous section .
we write @xmath194 for the negative logarithms of the model parameters .
consider any of the @xmath59 possible observations @xmath18 .
the quantity @xmath195 is the probability of making this particular observation , i.e. it is @xmath196 .
the quantity @xmath197 is the negative logarithm of the conditional probability @xmath198 where @xmath199 maximizes @xmath200 for the parameters @xmath201 . clearly , the function @xmath202 is piecewise - linear and concave on the logarithmic parameter space .
the domains of linearity of the function @xmath203 are the cones in the normal fan of the newton polytope of @xmath192 .
each maximal cone @xmath204 is indexed by the hidden data @xmath199 that maximizes @xmath205 for any of the parameters @xmath206 .
the hidden data @xmath199 which arise in this manner , for some choice of logarithmic parameters @xmath207 , are called the possible _ explanations _ of the observation @xmath18 .
for instance , for the hidden markov model of section 2 , the explanations are the viterbi sequences .
let us now vary the observations .
each logarithmic parameter vector @xmath208 defines an _ inference function _ @xmath209 from the set of observations to the set of explanations . for the hmm
, each inference function @xmath210 takes an observed sequence @xmath211 to the corresponding viterbi sequence @xmath199 .
there are @xmath212 such functions , but most of these are * not * inference functions .
for instance , consider the binary hmm of length three .
there are @xmath213 boolean functions @xmath214 , but , as we have seen at the end of section 2 , only @xmath75 of these are inference functions for the hmm . the inference functions @xmath209 of a graphical model @xmath28 are in bijection with the vertices of the newton polytope of the map @xmath28 .
the explanations @xmath215 for a fixed observation @xmath18 in a graphical model are in bijection with the vertices of the newton polytope of the polynomial @xmath192 . in applications of graphical models ,
the number @xmath25 of parameters and the number @xmath31 of values of the observed random variables is small and fixed , but the number @xmath29 of observed random variables is large .
recall that the model is the image of the map @xmath191 .
hence the dimension of the model remains fixed but the dimension of its ambient space grows exponentially in @xmath29 .
it is therefore algorithmically infeasible to compute the full tropical variety @xmath96 .
what we can do efficiently , however , is to compute the newton polytopes of the @xmath192 , or even the newton polytope of @xmath28 .
this allows us to glean information about the tropical variety from the domains of linearity of its `` coordinate functions '' @xmath203 .
our next goal is to derive an upper bound on the number of vertices of the newton polytopes .
[ smallpolytopebound ] consider graphical models @xmath28 whose number of parameters @xmath25 is fixed and whose number @xmath29 of observed random variables and number of edges @xmath9 varies .
( typically , @xmath9 is a linear function of @xmath29 ) .
then the number of vertices of the newton polytope @xmath216 of @xmath48 is bounded above by @xmath217 for many important families of graphical models , the number @xmath9 of edges is bounded by a linear function in terms of the number @xmath29 of observed nodes , and in those cases we can replace @xmath9 by @xmath29 . hence , for any given observation @xmath211 , the number of explanations grows polynomially in @xmath29 .
for instance , in the hidden markov model of section 2 we have @xmath218 , and a similar relationship holds in the tree model of section 5 .
[ hmmcomplexity ] for any fixed observation in the homogeneous hmm , the number of explanations is at most @xmath219 . if all random variables are binary then the upper bound @xmath220 holds .
the proof of theorem [ smallpolytopebound ] and corollary [ hmmcomplexity ] are derived from the following classical result on lattice polytopes due to andrews @xcite .
the necessary observation is that the newton polytope of @xmath48 is contained in the cube @xmath221^d$ ] , and the volume of this cube equals @xmath222 .
( andrews @xcite ) for every fixed integer @xmath25 there exists a constant @xmath223 such that the number of vertices of any lattice polytope @xmath162 in @xmath102 is bounded above by @xmath224 .
the newton polytope of the map @xmath28 was defined as the minkowski sum of the @xmath59 smaller newton polytopes in theorem [ smallpolytopebound ] . from this
we infer the following naive bound on its number of vertices .
the number of inference functions of a graphical model is at most @xmath225 , hence this number scales at most singly exponentially in the complexity @xmath226 of the graphical model . consider the homogeneous hmm on binary random variables .
each inference function is a boolean function @xmath227 , but not conversely .
the number of all boolean functions is @xmath228 , which grows doubly exponentially in @xmath29 .
however , the number of inference functions is at most @xmath229 . in practical applications of graphical models
, it may be infeasible to compute all ( singly - exponentially many ) inference functions .
nonetheless , we believe that important insight can be gained by computing and classifying the newton polytopes of graphical models @xmath28 on few random variables
. such a study would be the polyhedral analogue to the algebraic classification of @xcite .
on the other hand , for a fixed observation @xmath211 , the size of the newton polytope of @xmath192 grows polynomially with the size of the graphical model , and therefore there is hope that the polytopes can be computed efficiently .
despite the fact that the newton polytope of @xmath230 has polynomially many vertices in the size of the graphical model , the number of terms in @xmath230 grows exponentially .
this is a potential problem because the computation of the newton polytope requires inspecting these terms .
the following result states that , in fact , the convex hull computations scales with the running time of the sum - product algorithm , which for many models of interest scales polynomially with the size of the graphical model .
[ polyprop ] the newton polytopes of the polynomials @xmath48 can be computed recursively using the decomposition of @xmath48 according to the sum - product algorithm .
taken together , theorem 7 and proposition 11 say that * polytope propagation is an efficient algorithm for parametric inference with graphical models*. this statement is thesis ( c ) in our companion paper @xcite . in that paper , the sum - product algorithm and the polytope propagation algorithm are explained and analyzed in more detail .
we also demonstrate the practicality of our mathematical theory by explicitly computing ( and statistically interpreting ) various high - dimensional newton polytopes for graphical models that arise in biological sequence analysis .
we conclude by illustrating the concepts we have developed in the context of tree markov models . these are directed graphical models where the graph is a directed tree @xmath231 with observed random variables @xmath30 at the leaves .
the naive bayes model in section 3 is the special case where @xmath232 .
each edge @xmath193 has a different transition matrix @xmath233 $ ] .
we consider the general model in allman and rhodes @xcite , which means that the @xmath234 are arbitrary distinct @xmath235-matrices . in most applications ,
the transition matrices are from a special model family ( e.g. in phylogenetics these may be jukes - cantor model or the hasegawa - kishino - yano model ) .
as before , we relax the hypothesis that transition probabilities are non - negative and sum to @xmath44 .
hence the @xmath236 are distinct unknowns .
for simplicity we shall further assume that the tree @xmath231 is binary .
[ treeobservation ] the general markov model for the binary tree @xmath231 is the image of a map @xmath237 , where each coordinate of @xmath28 is a multilinear polynomial in the unknowns @xmath238 , @xmath193 edge of @xmath239 .
if we denote an edge between nodes @xmath240 and @xmath241 by @xmath242 and @xmath243 is the tree @xmath231 without the leaves , then the coordinate of the multilinear map @xmath28 indexed by an observed sequence @xmath244 can be written as follows : @xmath245 here @xmath246 ranges over all colorations @xmath247 of the nodes such that @xmath248 for all leaves @xmath241 .
our running example in this section is the binary tree in figure [ fig : tree ] with binary random variables @xmath249 . leaves . ]
-0.4 cm in this example , the coordinates of the multilinear map @xmath250 are given by the formula @xmath251 the prime ideal @xmath58 of polynomial invariants is generated by the @xmath160-subdeterminants of the matrix @xmath252 thus this particular model is the @xmath253 instance of the determinantal variety in proposition [ threebythree ] .
we generalize the determinantal presentation in this example by proposing the following explicit solution to problem 5 for arbitrary binary trees @xmath231 .
every edge of @xmath231 induces a _ split _ of the set of leaves @xmath254 , corresponding to the two connected components of the tree obtained by removing that edge . the unrooted tree underlying @xmath231 is uniquely determined by the set of these splits .
[ luckydeterminants ] the ideal @xmath58 of phylogenetic invariants of the general markov model for any binary tree @xmath231 on binary random variables is generated by the @xmath160-determinants of all two - dimensional matrices obtained by _ flattening _ the @xmath255-table @xmath256 according to the splits induced by the edges of @xmath231 .
we need to explain the meaning of the word `` flattening '' .
if @xmath257 is any split of the set @xmath258 then this refers to the @xmath259-matrix whose rows and columns are indexed by functions @xmath260 and @xmath261 respectively , and whose entries are the @xmath262 probabilities @xmath8 .
in december 2003 , allman and rhodes announced a proof of the set - theoretic version of our conjecture [ luckydeterminants ] .
what this means algebraically is that @xmath58 equals the radical of the ideal generated by the aforementioned @xmath160-determinants . in light of this progress
, we wish to offer also the following tropical version of conjecture [ luckydeterminants ] .
it would be very nice to show that proposition [ threebythree ] extends to this situation .
however , none of the remaining discussion in this section depends on these conjectures .
[ luckytropical ] the map @xmath28 is surjectively positive for @xmath263 . the tropical variety ( resp .
the positive tropical variety ) of the prime ideal @xmath58 coincides with the set of all @xmath264-tables @xmath265 whose flattenings along the splits of the tree @xmath231 have tropical rank ( resp .
barvinok rank ) at most @xmath163 .
the sum - product algorithm is used in practice to evaluate the polynomial ( [ treemap ] ) .
its running time is linear in @xmath29 , despite the fact that the number @xmath266 of terms in ( [ treemap ] ) grows exponentially .
this reduction in complexity is achieved by recursively grouping subsums .
for instance , ( [ treemapexample ] ) becomes @xmath267 the rule to remember is this : polynomials are evaluated recursively as sums of products of smaller polynomials .
this is the solution to problem 1 . for details on the tree case
see @xcite .
problem 2 is known in phylogeny as the _ joint ancestral reconstruction _
problem , which asks for the maximum a posteriori ancestral assignments @xmath268 given the observations @xmath269 at the leaves .
an efficient method for solving this problem appears in @xcite .
this method is nothing but the sum - product algorithm with ordinary arithmetic @xmath270 replaced by tropical arithmetic @xmath271 .
the @xmath211-coordinate of the tropicalization @xmath272 of the map ( [ treemap ] ) is @xmath273 this expression can be evaluated efficiently by the same scheme as before .
the rule now is this : piecewise - linear concave functions are evaluated recursively as minima of sums of smaller such functions . a simple example illustrating this rule is the tropicalization of ( [ treemapexample2 ] ) : @xmath274 where @xmath275 and similarly for @xmath276 .
we saw in section 4 that the number of vertices of the newton polytopes of the coordinate polynomials @xmath48 is critical for efficient parametric inference .
that number grows polynomially in @xmath29 if the number of parameters is fixed ( thanks to theorem [ smallpolytopebound ] ) but it may grow exponentially if the number of parameters is not bounded . for the general markov model on a tree @xmath231 , the growth will be exponential unless we restrict the number of parameters .
this can be done , for instance , by considering the _ homogeneous tree model _ where the transition matrices along all edges are identical : @xmath277 using theorem [ smallpolytopebound ] , we obtain the following result analogous to corollary [ hmmcomplexity ] .
the number of vertices of the newton polytope of any coordinate @xmath48 in the homogeneous tree model is bounded above by @xmath278 times a constant depending only on @xmath31 . for tree models which are used in applications , such as phylogenetics ,
the number of parameters is likely to be reduced even further . in such cases ,
the parametric joint ancestral reconstruction problem can be solved efficiently using the polytope propagation algorithm techniques in proposition [ polyprop ] .
the algebraic representation for graphical models with hidden variables leads naturally to an interpretation of a parameterized model as a point on an algebraic variety .
marginal probabilities are coordinates of points on the variety .
varieties can be tropicalized , and the statistical meaning is that the map probabilities ( calculated with logarithms of the parameters ) can be interpreted as coordinates of points on the positive part of the tropical variety . hence
, the tropical model is fundamental for understanding map probabilities .
although we have not addressed it in this paper , the logarithms of the marginal probabilities are coordinates of points on the _ amoeba _ @xcite of the model .
amoebas are likely to be important for understanding the geometry of maximum likelihood estimation .
the sum - product algorithm for graphical models is an efficient method for evaluating the coordinate polynomials of a graphical model .
this algorithm works in exactly the same way for classical arithmetic @xmath270 and for tropical arithmetic @xmath279 .
this means that the same method is used to evaluate coordinates of points on the variety and of points on the tropical variety .
an explanation for an observation @xmath18 is a vertex of the newton polytope of @xmath192 .
thus , the parametric inference problem is solved by finding the normal fans of the newton polytopes of the coordinate polynomials . for many important applications ,
the number of vertices of the polytopes is polynomial in the size of the graphical model .
the polytope propagation algorithm , which is a geometric analog of the sum - product algorithm , finds the newton polytopes , and is efficient when the sum - product algorithm is fast and the number of vertices on the newton polytopes is small .
an inference function for a graphical model is a function from the set of observations to the set of explanations which maximizes the a posteriori probabilities with respect to some choice of parameters .
inference functions correspond to vertices of the newton polytope of the map @xmath28 .
this polytope is much larger than the newton polytope of a single coordinate @xmath48 , so it can only be computed for small graphical models , but it has the advantage that it encodes the entire piecewise - linear geometry of the model . in a companion paper @xcite
, we show that polytope propagation is practical and useful in the important application of biological sequence analysis . in particular , existing parametric alignment methods @xcite
can be viewed as special cases of parametric inference for pair hidden markov models .
the computation of the newton polytopes is also useful for bayesian computations , where we have priors on the parameters and it is of interest to integrate over the maximal cones in the normal fan of the newton polytope @xcite .
lior pachter was supported in part by a grant from the nih ( r01-hg02362 - 02 ) .
bernd sturmfels was supported by a hewlett packard visiting research professorship 2003/2004 at msri berkeley and in part by the nsf ( dms-0200729 ) .
we are grateful to komei fukuda , michael joswig and kristian ranestad for their help in obtaining the computational results reported in section 2 .
26 g. andrews : a lower bound for the volume of strictly convex bodies with many boundary points , trans .
106 ( 1963 ) 270273 .
e. allman and j. rhodes : phylogenetic invariants for the general markov model of sequence mutation , mathematical biosciences , 186 ( 2003 ) 113144 .
s. aji and r. j. mceliece : the generalized distributive law , ieee transactions on information theory 46 ( 2000 ) 325343 .
p. baldi and s. brunak : bioinformatics . the machine learning approach . a bradford book . the mit press .
cambridge , massachusetts 1998 .
j. cavender and j. felsenstein : invariants of phylogenies in a simple case with discrete states , journal of classification 4 ( 1987 ) 5771
. j. cohen and u. rothblum : nonnegative ranks , decompositions , and factorizations of nonnegative matrices , linear algebra appl . 190 ( 1993 )
david cox , donal oshea and john little : ideals , varieties and algorithms , springer undergraduate texts in mathematics , 1996 .
m. develin , f. santos and b. sturmfels : on the tropical rank of a matrix , preprint , http://front.math.ucdavis.edu/math.co/0312114 .
m. develin : the space of @xmath24 points on a tree with @xmath29 leaves , preprint . .
r. durbin , s. eddy , a. krogh and g. mitchison : biological sequence analysis ( probabilistic models of proteins and nucleic acids ) , cambridge university press , 1998 .
d. fernndez - baca , t. sepplinen and g. slutzki : parametric multiple sequence alignment and phylogeny construction , in combinatorial pattern matching , lecture notes in computer science ( r. giancarlo and d. sankoff eds . ) , vol .
1848 , 2000 , 6882 .
garcia , m. stillman and b. sturmfels : algebraic geometry of bayesian networks , journal of symbolic computation , to appear , http://front.math.ucdavis.edu/math.ag/0301255 .
d. gusfield , k. balasubramanian , and d. naor : parametric optimization of sequence alignment , algorithmica 12 ( 1994 ) 312326 .
f. kschischang , b. frey , and h. a. loeliger : factor graphs and the sum - product algorithm , ieee trans .
theory 47 , feb 2001 , 498519 .
jordan and y. weiss : graphical models : probabilistic inference , in _ handbook of brain theory and neural networks , 2nd edition _ , m. arbib ( ed . ) , cambridge , ma , mit press , 2002
. t. pupko , i. peer , r. shamir , and d. graur : a fast algorithm for joint reconstruction of ancestral amino acid sequences , molecular biology and evolution 17 ( 2000 ) 890896 .
j. richter - gebert , b. sturmfels , and t. theobald : first steps in tropical geometry , in idempotent mathematics and mathematical physics ( eds .
litvinov and v.p .
maslov ) , american mathematical society , 2004 , preprint posted at http://front.math.ucdavis.edu/math.ag/0306366 .
l. pachter and b. sturmfels : parametric inference for biological sequence analysis , companion paper , submitted .
g. pistone , e. riccomagno , and h.p .
wynn : algebraic statistics : computational commutative algebra in statistics , chapman and hall , boca raton , florida , 2001 .
l. r. rabiner : a tutorial in hidden markov models and selected applications in speech recognition , proc . of the ieee , 77 ( 1989 ) 257286 .
d. speyer and l. williams : the tropical totally positive grassmannian , preprint , http://front.math.ucdavis.edu/math.co/0312297 . b. sturmfels : grbner bases and convex polytopes , university lecture series , vol . 8 , american mathematical society , 1996 . o. viro : what is an amoeba
? notices of the american math .
society 49 ( 2002 ) 916917 .
m. waterman , m. eggert and e. lander : parametric sequence comparisons , proc . natl . acad .
usa 89 ( 1992 ) 60906093 .
d. west : acyclic orientations of complete bipartite graphs , discrete math . 138 ( 1995 ) 393396 . | this paper presents a unified mathematical framework for inference in graphical models , building on the observation that graphical models are algebraic varieties . from this geometric viewpoint ,
observations generated from a model are coordinates of a point in the variety , and the sum - product algorithm is an efficient tool for evaluating specific coordinates .
the question addressed here is how the solutions to various inference problems depend on the model parameters .
the proposed answer is expressed in terms of tropical algebraic geometry .
a key role is played by the newton polytope of a statistical model .
our results are applied to the hidden markov model and to the general markov model on a binary tree . |
dust grains produce extinction and reddening of stellar light from the ultraviolet ( uv ) to the infrared ( ir ) ( draine 2003 ) . accurate determination of reddening to a star is vital for reliable derivation of its basic stellar parameters , such as effective temperature and distance .
constructing a 3d galactic extinction map plays an essential role in galactic astronomy , particularly in achieving the driving goals of the lamost spectroscopic survey of the galactic anti - center ( lss - gac ; liu et al .
this volume ) . the sloan digital sky survey ( sdss ; york et al .
2000 ) has delivered low - resolution spectra for about 0.7 m stars in its data release 9 ( dr9 ; ahn et al .
2012 ) . the lamost galactic surveys ( deng et al . 2012 and
this volume ) have obtained over 1 m stellar spectra and will obtain over 5 m in the next four years . with millions of stellar spectra , `` identical '' stars in different environments can be easily paired and compared , which opens great opportunities to a number of studies with the standard pair technique ( stecher 1965 ; massa , savage & fitzpatrick 1983 ) . by comparing the differences in photometric colors from large scale imaging surveys , such as the galaxy evolution explore ( galex ; martin et al .
2005 ) in the uv , the sdss in the optical , the two micro all sky survey ( 2mass ; skrutskie et al .
2006 ) in the near - ir and the wide - field infrared survey explorer ( wise ; wright et al .
2010 ) in the mid - ir , one can measure multi - band reddening for a large number of targets and constrain the reddening laws ( yuan , liu & xiang 2013 ) . by comparing the differences in normalized spectra ,
one can detect absorption features from the interstellar medium ( ism ) , particularly the diffuse interstellar bands ( dibs ; yuan & liu 2012 ) , as well as abnormal stellar absorption / emission lines from chemically particular or active stars .
such method has the advantages that it s straight - forward , model - free and applicable to the majority of stars . combining stellar distances from gaia ( perryman et al .
2001 ) or from spectro - photometry , one can further map the galactic extinction , extinction laws and dibs in 3d .
if by chance a diffuse nebula falls in the sightline of some targets , lines emitted by the nebula ( e.g. [ o iii ] @xmath04959 , 5007 ) will also be recorded and can be used to discover new faint emission line nebulae ( e.g. planetary nebulae and supernova remnants ) in the galaxy ( yuan & liu , submitted ) .
finally , the technique can also be used to perform photometric calibration for wide field imaging surveys ( yuan et al . in prep ) . in the following sections , we use the cases above to demonstrate the power of the standard pair technique when applied to large scale spectroscopic and photometric datasets including the sdss and on - going lamost surveys .
using star pairs selected from the sdss , and combining the sdss , galex , 2mass and wise photometry ranging from the far - uv to the mid - ir , yuan , liu & xiang ( 2013 ) have measured dust reddening in the @xmath1 and @xmath2 colors for thousands of galactic stars .
the measurements , together with the @xmath3 values given by schlegel et al .
( 1998 ) , allow us to derive the empirical reddening coefficients for those colors .
the results are compared with previous measurements and the predictions of a variety of galactic reddening laws .
we find that 1 ) the dust reddening map of schlegel et al .
( 1998 ) over - estimates @xmath3 by about 14% , consistent with the work of schlafly et al .
( 2010 ) and schlafly & finkbeiner ( 2011 ) ; and 2 ) all the new reddening coefficients , except those for @xmath4 and @xmath5 , prefer the @xmath6 = 3.1 fitzpatrick reddening law ( fitzpatrick 1999 ) rather than the @xmath6 = 3.1 ccm ( cardelli et al .
1989 ) and odonnell ( odonnell 1994 ) reddening laws . using the @xmath7-band extinction coefficient predicted by the @xmath6 = 3.1 fitzpatrick law and the observed reddening coefficients ,
we have deduced new extinction coefficients for the @xmath8 and @xmath9 passbands .
we recommend that the new reddening and extinction coefficients should be used in the future and an update of the fitzpatrick reddening law in the uv is probably necessary .
lss - gac has obtained about 0.8 m spectra of @xmath10 per pixel and basic stellar parameters for about 0.6 m spectra of @xmath11 per pixel ( liu et al .
this volume ) . with the same technique
, we measured dust reddening in the @xmath12 and @xmath13 colors for over 0.2 m stars from lss - gac .
the reddening coefficients for these colors relative to that for @xmath14 are consistent with the work of yuan , liu & xiang ( 2013 ) , as seen in fig.1 .
with spectro - photometric distances , we have constructed a preliminary 3d extinction map in the outer disk of the galaxy .
fig.2 shows @xmath3 as a function of distance from the sun and galactic longitude at @xmath15 , 0@xmath16 and @xmath17 . in spite of the limited sky coverage , distinct features , such as the perseus and outer arms , and effects of warps of the outer disk ,
are clearly visible . given the multi - band reddening of the stars , @xmath6 values are also estimated and their spatial variations are investigated in fig.3 .
a median value of @xmath6 = 3.2 is obtained , consistent with previous results .
no obvious spatial variations of @xmath6 are detected , indicating that dust properties do not change significantly in the outer disk .
the lss - gac will obtain low - resolution spectra and basic stellar parameters for a statistically complete sample of @xmath18 m stars in a large contiguous sky area ( @xmath19 , @xmath20 ) . for @xmath21 ,
the lss - gac plans to sample 1,000 stars per sq.deg . for @xmath22
, the sampling is doubled .
combining spectroscopic data from the lamost and sdss , photometric data from the galex , sdss , the xuyi schmidt telescope photometric survey of the galactic anti - center ( xstps - gac ; liu et al .
this volume ) , pan - starrs ( kaiser 2004 ) , 2mass and wise , and spectro - photometric distances and gaia parallaxes in the future , we will produce high spatial resolution ( about 10 arcmin ) , multi - band extinction maps in the galactic anti - center , and then study the distribution of dust and variations of extinction laws .
, @xmath23 , @xmath24 and @xmath13 colors versus that of @xmath14 for 225,422 stars from lss - gac that have xstps - gac , 2mass photometry and lamost spectral @xmath25 per pixel .
to avoid crowdness , only one - in - five stars are shown .
for comparison , relations obtained in yuan , liu & xiang ( 2013 ) are over - plotted .
, width=480 ] as a function of distance from the sun ( ranging from 0 10 kpc ) and galactic longitude ( ranging from 150@xmath16 210@xmath16 ) in the galactic outer disk . ,
width=480 ] ( left ) and its scattering ( right ) .
the x - axis and y - axis are galactic longitude ranging from 230@xmath16 120@xmath16 and latitude ranging from @xmath2640@xmath16 50@xmath16 , respectively .
, width=336 ]
dibs are weak absorption features detected in the spectra of reddened stars from the near uv to the near ir .
dibs have been discovered for almost a century , and to date over 400 dibs have been detected in galactic and extragalactic sources ( e.g. hobbs et al .
2008 , 2009 ) , but none of their carriers is identified ( herbig 1995 ; sarre 2006 ) .
the nature of dibs remains one of the most challenging problems in astronomical spectroscopy .
most recent work to identify and investigate the properties and carriers of dibs concentrates on high - resolution spectroscopy of a small number of selected sight - lines . using a template subtraction method based on the standard pair technique , yuan & liu ( 2012 )
have successfully identified the dibs @xmath05780 , 6283 in the sdss low - resolution spectra of a sample of about 2,000 stars and measured their strengths and radial velocities .
the sample is by far the largest ever assembled .
the targets span a large range of reddening , @xmath3 @xmath27 0.2 1.0 , and are distributed over a large sky area and involve a wide range of stellar parameters , confirming that the carriers of dibs are ubiquitous in the diffuse ism .
the sample is used to investigate relations between strengths of dibs and magnitudes of line - of - sight extinction , yielding results ( i.e. , @xmath28(5780 ) @xmath29 @xmath3 and @xmath28(6283 ) = 1.26 @xmath30 @xmath3 ) consistent with previous studies ( e.g. friedman et al . 2011 ) .
dib features have also been detected in the lamost spectra ( fig.4 ) of resolving power similar to that of sdss . in the commissioning spectra of an emission line star ,
even 9 dibs have been detected ( see fig.5 from yuan & liu 2012 ) .
detections of dibs towards hundreds of thousands of stars are expected with lamost .
the huge dib database will provide an unprecedented opportunity to study the demographical distribution of dibs . when combined with other data - sets
, it will enable us to address questions like : how the properties of dibs ( e.g. dib - dib , dib - extinction , dib - gas relations ) depend on local environment ( e.g. uv radiation field , @xmath6 , extinction in the fuv and the 2175 extinction bump ) ? where are their carriers formed ?
can they be formed in the circumstellar environments ? meanwhile , dibs and atomic absorption lines from the ism can act as good tracers to probe the distribution and properties of the ism and dust .
@xmath315780 , 6283 detected in the lamost spectra of three f / g dwarfs .
basic information of each star is listed at the bottom of each row . for each row ,
segments of the target and the scaled template spectra are shown in the left two panels .
the residuals are plotted in the right two panels , with the line fitting results over - plotted .
, width=384 ]
using @xmath27 1.7 m spectra from the sdss dr7 ( abazajian et al . 2009 ) , we have undertaken a systematic search for galactic planetary nebulae ( pne ) via detections of the [ o iii ] @xmath04959 , 5007 lines ( yuan & liu , submitted ) .
examples of the sdss spectra with well detected [ o iii ] @xmath04959 , 5007 lines are shown in fig.5 .
thanks to the excellent sensitivity of the sdss spectroscopic surveys , this is by far the deepest search for pne ever taken , reaching a surface brightness of the [ o iii ] @xmath315007 line s@xmath32 down to about 29.0 magnitude arcsec@xmath33 .
the search recovers 14 previously known pne in the galactic caps .
most of them are clearly visible on the sdss broad - band images owe to their high surface brightness . in total ,
about 60 new planetary nebula ( pn ) candidates are identified , including 7 probable candidates of multiple detections .
all the probable candidates are extremely large ( between 21 154arcmin ) and faint , located mostly in the low galactic latitude region with a kinematics similar to disk stars , confirming the presence of a significant population of previously undetected , large , nearby , highly evolved pne in the solar neighborhood .
four of the candidates have angular sizes between 84 154arcmin , and might well be the largest pne ever reported .
based on sky positions and kinematics , 12 of the possible candidates probably belong to the halo population .
if confirmed , they will double the numbers of known pne in the galactic halo .
most newly identified pn candidates are very faint , with s@xmath32 between 27.0 30.0 magnitude arcsec@xmath33 , and very challenging for previously employed techniques ( e.g. slitless spectroscopy , narrow - band imaging ) .
they greatly increase the number of faint pne and may well represent the `` missing '' pn population .
the results have demonstrated the power of large scale fiber spectroscopy in hunting for ultra - faint pne and other types of nebulae by detecting nebular emission lines .
the detection limits can be further increased by applying the same method which is used to detect dibs to the sdss and lamost stellar spectra . with millions of spectra from the sdss , lamost and other projects
, it will provide a statistically meaningful sample of ultra - faint and large pne as well as new supernova remnants to improve their censuses .
@xmath04959 , 5007 lines from galactic pne and pn candidates .
the png identification , sdss spectral i d , initial target type and redshift are labeled .
the fluxes are in unit of 10@xmath34ergs cm@xmath33 s@xmath36 @xmath36 .
, width=384 ]
uniform photometric calibration plays a central role in the large - scale imaging surveys , such as the sdss , the dark energy survey ( des ; the des collaboration 2005 ) , pan - stars and lsst ( ivezic et al .
2008 ) . the stellar locus regression ( slr ) method ( high et al . 2009 ) , adopted in des , can make one wholesale correction for differences in instrumental response , for atmospheric transparency , for atmospheric extinction , and for galactic extinction by adjusting the instrumental broadband colors of stars to bring them into accord with a universal stellar color - color locus , yielding calibrated colors accurate to a few percent .
the slr method assumes the standard stellar locus is universal , which is however not always true , due to varying stellar populations and extinction , especially in the galactic disk region .
it also requires a blue filter in addition to at least two of any other filters . to overcome the limitations above
, we propose a new method to perform photometric calibration using star pairs from large scale spectroscopic surveys ( yuan et al . in prep ) , such as lss - gac .
the star pairs here are composed of target and control stars from uncalibrated and calibrated fields , respectively .
this method requires that 1 ) extinction values of the targets are known , which can be from schlegel et al .
( 1998 ) or derived from existing photometric data ; 2 ) the reddening laws do not change in one field ; and 3 ) there are a few calibrated fields to obtain the intrinsic colors
. then it makes one wholesale correction by adjusting the instrumental colors of target stars to bring them into accord with their intrinsic colors , and obtain the reddening coefficient simultaneously .
if the accuracy of instrumental colors is about 1% and about 100 star pairs can be selected for each field , this method will yield a color calibration accuracy about a few mmag .
it is also useful in checking and improving calibration accuracies of existing surveys . | with modern large scale spectroscopic surveys , such as the sdss and lss - gac , galactic astronomy has entered the era of millions of stellar spectra . taking advantage of the huge spectroscopic database , we propose to use a `` standard pair '' technique to a ) estimate multi - band extinction towards sightlines of millions of stars ; b )
detect and measure the diffuse interstellar bands in hundreds of thousands sdss and lamost low - resolution spectra ; c ) search for extremely faint emission line nebulae in the galaxy ; and d ) perform photometric calibration for wide field imaging surveys . in this contribution
, we present some results of applying this technique to the sdss data , and report preliminary results from the lamost data . |
the recently reported planet candidate around the nearby m dwarf gj 581 ( * ? ? ? * hereafter v10 ) has generated much public enthusiasm and a similar amount of skepticism within part of the scientific community @xcite .
if confirmed , it will be the first planet potentially capable of hosting life as we know it now @xcite .
the possible existence of this planet was announced based on analysis of the new hires / keck precision rv measurements combined with harps / eso data published by @xcite ( hereafter m09 ) .
v10 reported that the candidate planet gj 581@xmath0 ( hereafter _ planet g _ ) has a minimum mass of @xmath2 @xmath3 and a period of 36.5 days .
the data for gj 581 contain the signal of at least 4 other low mass planets ( planets @xmath4 ) .
dynamical studies showed that their masses can not be larger than @xmath51.4 times the reported minimum masses ( m09 and v10 ) .
of particular interest to this work is gj 581d .
the planet has a period of 67 days and a previously reported eccentricity of @xmath6 ( m09 ) . using only the harps data ( m09 )
, @xcite noted that the eccentric orbit of gj 581@xmath1 could disguise the signal of a low mass companion at half its period , similar to the candidate planet @xmath0 now reported by v10 . both eccentric planets and pairs in or near 2:1 mean - motion resonance have been discovered by rv and transit surveys .
each case implies a very different dynamical history so the distinction is important in understanding evolution of planetary systems .
the available rv data of gj 581 is strongly affected by yearly aliasing ( * ? ? ? * hereafter df10 ) . in section [ sec : aliases ] , we explain this aliasing and demonstrate how the eccentricity of a planet can be confused with an additional planet at half its period . in section [ sec : degeneracy ] , we show that planet gj 581@xmath0 has an orbital frequency that aliases to the eccentric harmonic of planet @xmath1 and therefore could be an artifact of the sampling cadence . in section [ sec : analysis ] , we reevaluate the significance of planet @xmath0 in the presence of this degeneracy by allowing all the planetary orbits to be eccentric . in the same section
, we quantify the probability of confusing planet @xmath0 with the eccentric harmonic of planet @xmath1 . in our analysis
, we encountered some caveats about the reality and uniqueness of the reported signal . in section [ sec : caveats ] , we describe these caveats and their relation with the failure of recent bayesian studies to confirm the presence of planet @xmath0 @xcite . given that the 6-planet solution proposed by v10 is the most significant one that is also physically allowed , we conclude that the presence of planet @xmath0 remains supported by the data .
frequency will also create aliases .
bottom panel : power spectra of the residuals to a 4-planet fit with eccentricities preliminarily fixed at 0 ( bottom ) , a noiseless synthetic planet @xmath0 ( center ) , and a noiseless synthetic signal at the frequency of the eccentric harmonic ( top ) .
the aliases are marked with symbols corresponding to the upper panel .
, width=453 ] the sampling cadence of a periodic signal can alter its apparent period .
an _ alias _ is a spurious periodicity generated by finite sampling of a real signal . as an extreme example , if a function of period @xmath7 is measured at regular intervals @xmath8 , it will appear constant to the observer . in the case of even sampling ,
the maximum frequency that can be probed without ambiguity is half the sampling frequency ( the nyquist frequency ) .
however , this frequency can not be unambiguously determined in the case of uneven sampling @xcite .
instead , one can identify characteristic sampling frequencies by computing the _ spectral window function _ @xmath9 , which depends only on the observation instants .
the modulus of @xmath9 ranges from 0 to 1 and peaks at the sampling frequencies that cause the most severe aliases . for a real signal of frequency @xmath10 and a characteristic sampling frequency @xmath11
, aliases appear at @xmath12 .
figure [ fig : window ] shows the window function including all the harps+hires data available for gj 581 ( m09 and v10 ) .
optical astronomical measurements occur only at night and only during the season when the star is observable .
the night time cadence introduces two sampling frequencies close to 1 day@xmath13 ( the solar and sidereal day , top right panel ) , and the seasonal availability generates a 1 year peak ( top left panel ) .
when the signal to noise is low , coherent addition of noise can produce higher power at the aliases than at the real period ( see df10 ) .
since the announcement of the first planet candidates around gj 581 by @xcite , the yearly alias has significantly affected the determination of the orbital periods of these planets .
for example , the first period proposed for the super - earth gj 581d was 84 days @xcite , a yearly alias of the currently accepted period of 67 days . .
bottom : periodograms of the same datasets , fixing @xmath14 at 0 .
the left panels correspond to synthetic data containg the six planets announced by v10 .
the right panels are periodograms of the residuals of actual data
. the 36.5 day signal will be missed if the eccentricity of planet @xmath1 is prematurely adjusted ( red question marks ) .
, width=491 ] the other ingredient to the degeneracy is that the rv signal due to a planet s keplerian motion can be expressed as a series in powers of its orbital eccentricity , @xmath15 . to first order in eccentricity
, a keplerian rv signal of amplitude @xmath16 and period @xmath7 can be written as @xmath17 where @xmath18 and @xmath19 are functions of the initial mean anomaly and the argument of the periastron , and @xmath20 is a constant offset that can be different for each instrument .
[ eq : rv ] implies that in a system with a planet of period @xmath7 , an additional body with period @xmath21 is statistically indistinguishable unless the next term in the expansion can be measured .
we call the term proportional to @xmath22 the `` first eccentric harmonic . '' due to statistical uncertainties , finite sampling , and aliasing , two planets can be misinterpreted as a single eccentric planet even when their period ratio is not exactly 2 .
as shown by m09 , the period of planet @xmath1 is @xmath23 days .
therefore , the first eccentric harmonic of this planet has a period of @xmath24 days ( @xmath25 day@xmath13 ) .
the period of the newly found planet @xmath0 is @xmath26 days , and its aliases should appear at @xmath27 and @xmath28 days ( @xmath29 day@xmath30yr@xmath31 ) . because a yearly alias of planet @xmath0 falls very close to the eccentric harmonic of planet @xmath1 , planet @xmath0 s signal will be partially absorbed by the eccentricity of planet @xmath1 . to illustrate this , we carry out the following experiment .
we generate a synthetic dataset , the exact 6-planet solution from v10 ( all circular orbits ) sampled at the times of the harps and hires datasets and with gaussian noise added ( 1.9 and 2.7 m / s respectively ) . then , we compute the periodogram of the residuals of a 4-planet fit to this synthetic dataset ( fig . [ fig : aliasing ] ) .
even though planet @xmath0 is included in the simulated data , its peak will disappear if planet @xmath1 is allowed to be eccentric ( top left ) . on the other hand
, planet @xmath0 will appear prominently if the eccentricity of planet @xmath1 is preliminarily fixed at 0 ( bottom left ) .
the same behavior is observed for the real data ( right panels ) . as a general rule in modeling rv data sets ,
it is safer to search for significant periodicities first ( i.e. preliminarily fixing the eccentricities at 0 ) and defer determining which model is preferred to a later stage ( i.e. an eccentric planet vs. a two planet model ) . in df10
, a test was developed to qualitatively assess which period is more likely in the presence of signal aliasing .
the test consists of the following steps : ( 1 ) generate noiseless synthetic datasets of the signals under study , ( 2 ) compute their power spectra at the regions where the most prominent features occur ( the nominal period and strongest aliases ) , and ( 3 ) compare them with the power spectrum of the real data ( i.e. for gj 581 , the residuals to the 4-planet fit with eccentricities preliminarily fixed at 0 ) . by _ power spectrum _ we mean the amplitude of the sinusoidal function that best fits the data at each frequency ( see df10 for further details ) .
we generate a synthetic dataset for each of the two candidate periods ( 33.5 days and 36.5 days ) .
the bottom panels of figure [ fig : window ] show the power spectra of the real and the synthetic datasets at the nominal periods and their most prominent aliases . according to @xcite , if there is a satisfactory match between an observed spectrum and a noise - free spectrum of period @xmath7 , then @xmath7 is the true period " . while the eccentric harmonic ( 33.5 days ) and its aliases fail to mimic the features observed in the real data ( especially the daily aliases in the central panels ) , the candidate signal at 36.5 days does a fairly good job of reproducing most of them .
random noise modifies the power balance between peaks , so one should understand this as a first qualitative assessment .
the probability of confusing planet @xmath0 with the eccentric harmonic of planet @xmath1 is quantified in section [ sec : confusion ] .
in order to indentify promising low - amplitide periodicities , first we sequentially subtract the four most significant signals that coincide with those reported by m09 ( using the systemic interface @xcite ) . as discussed in section [ sec : degeneracy ] , we preliminarily fix the eccentricities at 0 .
all parameters are refined each time a planet is added .
the periodogram @xcite of the residuals to the 4-planet model is then computed .
both the 36.5 and 433 day signals appear prominently , see bottom right panel of fig.[fig : aliasing ] .
a _ detection _ false alarm probability ( fap ) is obtained through a monte carlo approach using 10@xmath32 synthetic realizations of the data .
each synthetic dataset is created by scrambling the residuals to the four planet fit , keeping the same observing epochs and instrument membership .
the detection fap is the fraction of times we obtain a signal with a higher power than the one in the real residuals .
we find that both _ detection _
fap are below 0.5% . after subtracting these two signals ( 36.5 and 433 days ) ,
no additional peaks are found with a detection fap under 1% .
these fap are for detection of sinusoid like signals only .
the actual significances for the proposed new candidates are computed in the next section .
recently , @xcite and @xcite indicated that planet @xmath0 is not robustly confirmed when bayesian analysis is applied . to address this issue
, we consider in this section the general case where the orbits of the already - detected planets are allowed to be eccentric and compute the definitive fap analytically using a classical frequentist approach . because the uncertainties in rv measurements are difficult to quantify , the @xmath33 statistic can not be used to evaluate the goodness - of - fit in an absolute sense @xcite .
however , the @xmath33 can still be used in a differential way , i.e. , to determine whether or not the addition of a planet is justified given the improvement of the @xmath34 ( @xmath35 are the residuals with respect to a proposed model and @xmath36 are the uncertainties ) . in this respect
, we apply a version of the fisher f - ratio test proposed by @xcite that allows the addition of any number of free parameters to the null hypothesis and to the model being tested .
this requires the computation of the f - ratio as @xmath37 where @xmath38 and @xmath39 are the number of free parameters of the null hypothesis and the model to be tested respectively .
the f ratio follows an f probability distribution with @xmath40 and @xmath41 degrees of freedom @xcite .
the null hypothesis includes the m09 planets ( planets @xmath42 ) and planet @xmath43 , all with fully keplerian orbits .
the @xmath44 is obtained by adding a new planet on a @xmath536.5 day circular orbit ( requiring two extra parameters ) and readjusting all the free parameters . using the values in the two last columns of table [ tab : stats ] , the obtained probability of improvement - by - chance at that particular frequency is @xmath45 .
now one has to consider the number @xmath46 of independent frequencies that could also generate a spurious improvement . according to v10 ,
m is @xmath47 for the harps+hires combined dataset .
therefore , the definitive fap for planet @xmath0 becomes fap@xmath48@xcite and amounts to @xmath49 . for completeness
, we add the fap of planet @xmath43 ( @xmath50 ) to table [ tab : stats ] , using the keplerian 4-planet solution ( m09 ) as the null hypothesis . the f - ratio test can also be used to evaluate the significance of the eccentricity compared to a circular orbit ( the null hypothesis ) .
we find that the eccentricities of the larger amplitude planets ( i.e. @xmath51 and @xmath15 ) are well - constrained but compatible with @xmath52 .
for the low amplitude candidates ( @xmath43 and @xmath0 ) , the eccentricities are unconstrained . table [ tab : solution ] provides the full orbital solution . at this point , we have assessed the fap of the planet detection .
now we need to quantify the probability that the eccentric harmonic of planet @xmath1 could inflate the signal at 36.5 days by an unfortunate combination of random errors .
we generate @xmath53 synthetic datasets by scrambling the residuals to the 6-planet fit and injecting a signal at 33.5 days ( amplitude @xmath54 m / s ) .
the periodogram of each dataset is then computed .
we define the _ probability of confusion _ as the number of times the period of the highest peak does not lie within 1 day of the injected signal , divided by the total number of trials . for the @xmath55 days
, we find this probability is @xmath56 . in @xmath57% of trials ,
the highest peak is found near the @xmath26 days alias . performing the same test by injecting the 36.5 days signal
, we obtain a @xmath58 probability of confusion ( with @xmath59 of trials corresponding to the 33.5 days alias ) . as a final check ,
we repeat the same experiment using random noise instead of scrambling the residuals ( @xmath60 1.8 m / s for harps , and @xmath60 3.0 m / s for hires ) . similarly , low confusion rates ( @xmath5 0.5% ) are obtained . in every case , the probabilities are below @xmath61 , supporting the conclusion of the previous section : _ the probability of confusing planet @xmath0 with the eccentric harmonic of planet @xmath1 is very low_. as a final check , we apply the same test using only the harps residuals . in more than @xmath62 of the trials ,
the highest peak is none of the injected signals , confirming that the data in m09 was not sensitive enough to distinguish between cases .
lccccc planets included & bced & bcde(f ) & & bcdef & bcdef(g ) + eccentric orbits & 4 & 4 & & 5 & 5 + circular orbits & 0 & 1 & & 0 & 1 + eccentricity + of planet @xmath1 & @xmath63 & @xmath64 & & @xmath64 & @xmath65 + + rms ( m / s ) & 2.32 & 2.21 & & 2.20 & 2.10 + free parameters & 22 & 24 & & 27 & 29 + @xmath66 & 241 & 241 & & 241 & 241 + @xmath33 & 707.7 & 611.8 & & 602.5 & 524.8 + + f ratio & & 17.0 & & & 15.69 + fap & & 0.03% & & & 0.11% + [ tab : stats ] during our analysis , we encountered a number of caveats that will require further investigation .
recent analyses by other authors using bayesian methods ( i.e. * ? ? ?
* ; * ? ? ?
* ) seem to contradict the conclusions reached here .
thus we discuss some of these caveats and possible lines of inquiry .
* significance of spurious periodicities .
* we found a number of alternative periodicities for planets @xmath43 and @xmath0 that yield alternative 6-planet fits with similar significance to the solution announced by v10 .
potential periods for planets @xmath43 and @xmath0 include 25.0 , 26.8 , 30.6 , 47.8 , 54.6 , 59.3 , 71.4 , 76.4 and 97.9 days .
these alternative solutions were found by computing a two - dimensional periodogram ( i.e. two periodic signals are adjusted simultaneously on a grid ) of the residuals of the 4-planet circular solution and refining all keplerian parameters around the areas of lowest @xmath33 @xcite .
however , a case - by - case investigation indicated that all correspond to orbital configurations with high eccentricities that suffer from several orbital crossings , making them dynamically unstable on very short timescales . to ultimately decide if such solutions were acceptable , we fixed the involved eccentricities to lower values , adjusting all other parameters in the process .
the resulting orbital fits were poorer than the 6-planet solution from v10 ( even assuming circular orbits for planets @xmath43 and @xmath0 ) .
the significance of these unphysical solutions raises the caveat that planet @xmath0 could be a physically - possible but spurious signal . as discussed by @xcite
, bayesian analysis methods do not yet apply down - weighting of unphysical configurations .
because such orbits have high significance , they will be oversampled , downgrading the likelihood of physically allowed orbits .
this may explain the apparent contradiction of the recent bayesian analysis with our conclusions .
* hidden planets*. while planets @xmath67 and @xmath68 are detectable in both datasets , planets @xmath1 and @xmath15 are not obvious in the hires data alone ( as also noted by v10 ) .
this indicates that some signals may be consistent with a given dataset , yet not independently detectable due to sampling issues . as a similar case ,
gj 876@xmath15 was a strong signal in the hires dataset yet undetectable with harps @xcite .
because the completeness of on - going rv surveys can be affected as a result , this caveat requires further investigation .
* noise unknowns*. we have chosen statistical tools that are as insensitive as possible to assumptions about noise : the f - ratio test and monte carlo scrambling of the residuals .
still , we recognize that it is not fully understood how systematic effects ( e.g. stellar jitter ) impact the sensitivity to low amplitude signals .
a bayesian approach with a _ noise parameter _ has been proposed to solve this problem @xcite
. however , we think that further tests are required to assess the sensitivity of bayesian methods in the presence of unknown systematic noise and low amplitude signals .
the eccentric harmonic of planet @xmath1 coincides with a yearly alias of the newly reported planet @xmath0 , meaning that both signals are correlated and that a premature keplerian fit to planet @xmath1 prevents the detection of planet @xmath0 .
we have found a number of unphysical solutions that fit the data just as well .
still , the proposed pair of planets @xmath43 and @xmath0 remains the only physically viable solution that significantly improves the 4-planet fit .
thus we are compelled to conclude that the presence of gj 581@xmath0 is well supported by the available data .
the ultimate confirmation will require additional rv measurements and reanalysis of the data in a very convincing way . to mitigate yearly aliasing , we encourage observations at more extreme parallax factors .
we end with two cautionary notes .
first , whether or not planet @xmath0 exists , the same cadence issues may be present in other datasets .
we have found that significant periodicities in one dataset do not appear in another and that several unphysical models can appear significant .
if future observations rule out the existence of planet @xmath0 , the fact that it passes the standard , widely - used statistical tests would bode ill for other low - amplitude planet candidates .
we remark that statistical significance tests can be very sensitive to assumptions by the authors ( including this work ) .
bayesian methods may provide stricter confidence level estimates but additional testing is required to ensure that they are not over - conservative in the low signal to noise regime @xcite .
bayesian methods may also need to demonstrate how their conclusions are impacted by sampling issues and the inclusion of dynamical constrains in the likelihood sampling process .
second , the eccentricity of a long period giant planet may cause spurious low amplitude signals mimicking habitable planets around sun - like stars . consider jupiter as an example
: its eccentric harmonic would have a period of 5.9 years ( or f = 0.00046162 days@xmath13 ) .
the corresponding yearly aliases would be at @xmath690.0027378 days @xmath70 , giving candidate signals at 439 and 313 days .
complementarily , a genuine planet candidate can be missed if the eccentricity of an outer planet is adjusted prematurely .
aliasing tests on the eccentric harmonics of detected planets might be necessary in future claims of very low - amplitude signals .
* acknowledgements * this work was funded by the carnegie fellowship postdoctoral program and the national science foundation graduate research fellowship .
we thank p. butler , and s. vogt for early access to the rv data .
we thank a. boss , j. chanam , j. dunlap , d. fabrycky , n. haghighipour , m. lopez - morales , and n. moskovitz for useful comments and discussions .
lrrrr @xmath7 [ days ] & 5.368483 ( 10@xmath71 ) & 12.91750 ( 0.002 ) & 66.845 ( 0.09 ) & 3.1485 ( 2 10@xmath72 ) + @xmath15 & 10@xmath72(10@xmath73)@xmath74 & 0.10 ( 0.05)@xmath74 & 0.11 ( 0.08)@xmath74 & 0.18 ( 0.08)@xmath74 + @xmath16 [ ms@xmath13 ] & 12.49 ( 0.15 ) & 3.40 ( 0.18 ) & 1.90 ( 0.15 ) & 1.75 ( 0.14 ) + @xmath76 [ deg ] & 87.9 ( 10 ) & 190.6 ( 7 ) & 61.5 ( 8) & 104.2 ( 7 ) + @xmath77 [ deg ] & 190.3 ( 8) & 193.3 ( 5 ) & 353.9 ( 7 ) & 144.7 ( 5 ) + + @xmath78 [ m@xmath79 & 15.61 ( 0.16 ) & 5.71 ( 0.23 ) & 5.4 ( 0.21 ) & 1.805 ( 0.16 ) + @xmath80 [ au ] & 0.041 & 0.073 & 0.218 & 0.028 + + new planets ( v10 ) & f & @xmath0 + + @xmath7 [ days ] & 440.7 ( 2.0 ) & 36.53 ( 0.32 ) + @xmath15 & 0.17@xmath81 & 0.19@xmath81 + @xmath16 [ ms@xmath13 ] & 1.18 ( 0.13 ) & 1.34 ( 0.17 ) + @xmath76 [ deg ] & 330.8@xmath81 & 68.9@xmath81 + @xmath77 [ deg ] & 183.24 ( 10 ) & 248.1 ( 10 ) + + @xmath78 [ m@xmath79 & 6.28 ( 0.65 ) & 3.18 ( 0.4 ) + @xmath80 [ au ] & 0.767 & 0.146 + + @xmath82 [ ms@xmath13 ] & -2.22 ( 0.2 ) + @xmath83 [ ms@xmath13 ] & 0.934 ( 0.2 ) + @xmath33 & 520.5 + rms [ m / s ] & 2.093 + [ tab : solution ] | the radial velocity ( rv ) method for detecting extrasolar planets has been the most successful to date .
the rv signal imprinted by a few earth - mass planet around a cool star is at the limit of the typical single measurement uncertainty obtained using state - of - the - art spectrographs .
this requires relying on statistics in order to unearth signals buried below noise .
artifacts introduced by observing cadences can produce spurious signals or mask genuine planets that should be easily detected otherwise .
here we discuss a particularly confusing statistical degeneracy resulting from the yearly aliasing of the first eccentric harmonic of an already - detected planet .
this problem came sharply into focus after the recent announcement of the detection of a 3.1 earth mass planet candidate in the habitable zone of the nearby low mass star gj 581 .
the orbital period of the new candidate planet ( gj 581@xmath0 ) corresponds to an alias of the first eccentric harmonic of a previously reported planet , gj 581@xmath1 . although the star is stable , the combination of the observing cadence and the presence of multiple planets can cause period misinterpretations . in this work , we determine whether the detection of gj 581@xmath0 is justified given this degeneracy .
we also discuss the implications of our analysis for the recent bayesian studies of the same dataset , which failed to confirm the existence of the new planet . performing a number of statistical tests ,
we show that , despite some caveats , the existence of gj 581@xmath0 remains the most likely orbital solution to the currently available rv data . |
in biological sequence analysis , probability distributions over finite ( @xmath0-dimensional ) sequences of symbols , representing nucleotides or amino acids , play a major role .
they specify the probability of a sequence belonging to a specified family , and are usually generated by markov chains .
these include the stochastic finite - state moore machines called hidden markov models ( hmms ) ; or infinite - state markov chains such as stochastic push - down automata ( spdas ) . by computing the most probable path through the markov chain
, one can answer such questions as `` what hidden ( e.g. , phylogenetic ) structure does a sequence have ? '' , and `` what secondary structure will a sequence give rise to ? '' .
the number of markov model parameters should ideally be kept to a minimum , to facilitate parameter estimation and model validation .
the a priori modeling of an rna sequence family is considered here . due to watson
crick base pairing , a recursively structured rna sequence will fold , and display secondary structure . to model stochastically both pairings and runs of unpaired bases ( which form loops and bulges ) , results from a subfield of formal language theory , the _ structure theory of weighted strings _ @xcite ( each string being weighted by an element of a specified ` semiring ' such as @xmath1 ) , are reviewed and employed in stochastic model construction . in section [ sec :
duration ] , _ duration modeling _ is discussed : the modeling of a probability distribution on ` runs ' , i.e. , on the natural numbers @xmath2 .
a non - rna biological example is the modeling and prediction of cpg islands in a dna sequence .
a sequence may flip between cpg and non - cpg states , with distinct hmms for generation of symbols in @xmath3 . for ease of hmm parameter estimation , and for finding the most probable parse , or path through the model ( e.g. , by the viterbi algorithm ) , the length of each cpg island and non - cpg region should be modeled in a markovian way , as the first hitting time in a finite - state markov chain .
that is , on @xmath4 , the set of possible lengths , it should have a _ phase - type distribution _ @xcite .
there is a theorem of the author s on such distributions @xcite , which grew out of results on positively weighted _ regular _ sequences @xcite .
it says that without loss of generality , the structure of the markov chain can be greatly restricted : its ` cyclicity ' can be required to be at most @xmath5 .
this has implications for hmm parametrization .
the generating function @xmath6 of a phase - type ( ph ) distribution on @xmath7 ( which is a normalized @xmath8-weighted regular language over a @xmath0-letter alphabet ) is a _ rational _ function of @xmath9 .
going beyond regular languages to the context - free case yields an _ algebraic _ generating function : one of several variables , if each type of letter in the sequence is separately kept track of . in rna secondary structure prediction , stochastic context - free grammars ( scfgs ) , usually in chomsky normal form , have been used @xcite .
they tend to be complicated ; if the grammar has @xmath10 non - terminal symbols , then it may have @xmath11 transition probabilities , which must be estimated from training sequences @xcite .
what is needed is a class of scfgs with ( i ) restricted internal structure , ( ii ) equivalent modeling power , and ( iii ) computationally convenient parametrization .
finding such a class of models is a hard problem : even on the level of @xmath0-letter - alphabet ( i.e. , univariate ) generating functions , it involves the constructive theory of algebraic functions . in section
[ sec : algebraic ] , as a small step toward solving this problem , it is pointed out that there is a class of probability distributions on @xmath7 with generating functions ( i.e. , @xmath9-transforms ) that are algebraic and non - rational , which can be conveniently parametrized .
this is the class of algebraic _
hypergeometric distributions_. e.g. , the @xmath2-valued random variable @xmath12 could satisfy @xmath13 , where @xmath14 is gauss s ( parametrized ) hypergeometric function .
if @xmath15 are suitably chosen , @xmath16 will be a probability density function with an algebraic @xmath9-transform .
algebraic hypergeometric probability densities satisfy nice recurrence relations , and scfg interpretations for them can be worked out .
a more general approach toward solving the above problem , not restricted to the case of a @xmath0-letter alphabet , employs scfgs with a two - level structure . in effect , these are scfgs wrapped around hmms .
the following is an illustration .
a probabilistically weighted dyck language over the alphabet @xmath17 , i.e. , a distribution over the words in @xmath18 that comprise nested @xmath19@xmath20 pairs , is generated from a symbol @xmath21 by repeated application of the production rule @xmath22 .
the probabilities @xmath23 sum to @xmath0 .
if each of @xmath24 in turn represents a weighted _ regular _ language over some alphabet @xmath25 ( e.g. , a ph - distribution if @xmath25 has only one letter ) , then the resulting distribution over words in @xmath26 comes from a scfg with the stated two - level structure .
this setup is familiar from ( unweighted ) language theory applied to compilation : the top - level structure of a program is specified as a word in a context - free language , and islands of low - level structure ( e.g. , identifier names and arithmetic literals ) as words in regular languages . in section [ sec : modeling ]
, it is indicated how the idea of a scfg wrapped around hmms can be applied to rna structure prediction : initially , to the parametric stochastic modeling , in a given sequence family , of the recursive primary structure that induces secondary folding .
the goal is parameter estimation and model validation , by comparison with data on real rna sequences .
knudsen and hein @xcite and nebel @xcite have worked on this , using dyck - like languages , but stochastic modeling using distinct scfg and hmm levels is a significant advance . on the level of primary rna structure , paired nucleotides will make up a subsequence of the full nucleotide sequence , and must constitute a dyck word , for simplicity written as a word over @xmath17 .
a distribution over the infinite family of such dyck words is determined by the above stochastic production rule , the parameters @xmath27 in which are specific to the family being modeled .
the production rule for _ full _ sequences , including unpaired nucleotides , will have not @xmath28 , @xmath29 , @xmath30 , @xmath31 on its right - hand side , but rather @xmath32 , @xmath33 , @xmath34 , @xmath35 , where each @xmath36 expands to a ` run ' of unpaired nucleotides . if the four nucleotides are treated as equally likely in this context , each @xmath36 will be a stochastic language over a @xmath0-letter alphabet , and the length of each run is reasonably modeled as having a ph - distribution .
the ph class includes geometric distributions , but is more general .
the overall scfg is obtained by wrapping the dyck scfg around the finite - state markov chains that yield the ph - distributions . from a given family of rna sequences ,
dyck scfg parameters can be estimated , e.g. , by the standard inside outside algorithm @xcite ; and then hmm parameters ( i.e. , ph - distribution parameters ) can be estimated separately . by employing a large enough class of ph - distributions ,
it should be possible to produce a better fit to data on secondary structure than were obtained from the few - parameter models of knudsen and hein @xcite and nebel @xcite .
once the family has been modeled , the most likely parse tree for any new rna sequence in the family can be computed by maximum a posteriori estimation , using the cyk algorithm @xcite .
the sequence is predicted to have the secondary structure represented by that parse tree .
since loops and bulges in rna secondary structure comprise runs of unpaired nucleotides , they can be modeled without taking long - range covariation into account .
the appropriate stochastic model is an hmm _ with absorption _ , since the accurate modeling of run lengths is a goal .
any such hmm will specify a probability distribution on the set of finite strings @xmath26 , where @xmath37 is the alphabet set , and long words are exponentially unlikely .
there should be little change in the nucleotide distribution along typical runs , so the distribution of the string length @xmath38 is what is important .
the time @xmath12 to reach a final ( absorbing ) state in an irreducible discrete - time markov chain on a state space @xmath39 , with a transition matrix @xmath40 that is _ substochastic _
( i.e. , @xmath41 ) , and an initial state vector @xmath42 that is also substochastic ( i.e. , @xmath43 ) , is said to have a discrete ph - distribution .
the substochasticity of @xmath44 and @xmath45 expresses the absorption of probability , since they can be extended to a larger state space @xmath46 , on which they will be stochastic .
the added state @xmath47 is absorbing .
there is a close connection between ph - distributions and finite automata theory , in particular the theory of rational series over semirings @xcite . if @xmath48 is a semiring ( a set having binary addition and multiplication operations , @xmath49 and @xmath50 , each with an associated identity element ; but not necessarily having a unary negation operation ) , then an @xmath48-_rational sequence _
, @xmath51 , is a sequence of the form @xmath52 $ ] , where for some @xmath53 , @xmath54 and @xmath55 .
it is an @xmath48-weighted regular language over a @xmath0-letter alphabet .
semirings of interest here include @xmath56 , @xmath57 , and the boolean semiring @xmath58 .
[ thm : normalization ] any ph - distribution on @xmath2 is an @xmath8-rational sequence .
any _ summable _
@xmath8-rational sequence , if normalized to have unit sum , becomes a ph - distribution .
if @xmath38 is ph - distributed , it is useful to focus on its @xmath9-transform , i.e. , @xmath59=\sum_{n=0}^\infty z^n\,pr(\tau = n)$ ] .
this will be a rational function , in @xmath60 .
if the distribution is _ finitely supported _ , it will be a polynomial , in @xmath61 $ ] .
[ thm:2 ] any ph - distribution on @xmath2 can be generated from finitely supported distributions by repeated applications of ( i ) the binary operation of mixture , i.e. , @xmath62 , where @xmath63 , ( ii ) the binary operation of convolution , i.e. , @xmath64 , and ( iii ) the unary ` geometric mixture ' operation , i.e. , @xmath65 , where @xmath63 .
this is a variant of the kleene
schtzenberger theorem on the @xmath48-rational series associated to @xmath48-finite automata @xcite .
the boolean ( @xmath66 ) case of their theorem is familiar from formal language theory : it says that any regular language over a finite alphabet can be generated from _
languages by repeated applications of ( i ) union , ( ii ) concatenation , and ( iii ) the so - called kleene star operation .
just as in formal language theory , the third operation of theorem [ thm:2 ] can be implemented on the automaton level by adding ` loopback ' , or cycle - inducing , transitions from final state(s ) back to initial state(s ) .
[ thm:3 ] the unary binary computation tree leading to any ph - distribution on @xmath2 , the leaves of which are finitely supported distributions , can be required without loss of generality to have at most @xmath5 unary ` geometric mixture ' nodes along the path extending to the root from any leaf .
that is , those operations do not need to be more than doubly nested .
this is a normalized , or ` stochastic ' , version of a result on the representation of @xmath8-rational sequences @xcite .
results of this type are strongly semiring - dependent .
it is not difficult to see that in the cases @xmath67 and @xmath68 , the analogue of the number ` @xmath5 ' is ` @xmath0 ' .
( this is because , e.g. , a @xmath68-rational sequence is simply an sequence in @xmath69 that is eventually periodic . )
the proof of theorem [ thm:3 ] is an explicit construction , which respects positivity constraints at each stage . that is , the construction solves the _ representation problem _ for univariate ph - distributions , which has strong connections to the positive realization problem in control theory @xcite .
what the theorem says , since operations of types ( i),(ii),(iii ) correspond to parallel composition , serial composition , and cyclic iteration of markov chains , is that any ph - distribution arises without loss of generality from a markov chain in which cycles of states are nested at most @xmath5 deep .
that is , the chain may include cycles , and cycles within cycles , but not cycles within cycles within cycles .
so for modeling purposes , the chain transition matrix @xmath44 may be taken to have a highly restricted structure . a completely connected transition graph on a state space of size
@xmath70 would have @xmath71 possible transitions , and would be unnecessarily general when @xmath70 is large .
unpaired nucleotide run lengths in rna are naturally modeled as having ( discrete - time ) ph distributions because of the close connection with hmms , and the consequent ease of parameter estimation .
however , the class of ph distributions is so versatile that it would be useful in this context , regardless .
discrete ph distributions include geometric and negative binomial distributions , and are dense ( in a suitable sense ) in the class of distributions @xmath72 , @xmath73 , which have leading - order geometric falloff as @xmath74 .
any ph distribution on @xmath2 has a @xmath9-transform @xmath59 $ ] that is rational in the conventional sense ; equivalently , it must satisfy a finite - depth recurrence relation of the form @xmath75 .
in fact , any probability distribution on @xmath2 with ( i ) a rational @xmath9-transform @xmath6 , and ( ii ) the property that the pole which @xmath6 necessarily has at @xmath76 is the _ only _ pole on the circle @xmath77 , is necessarily a ph distribution @xcite .
this is a sort of converse of the perron
frobenius theorem . however , there are distributions on @xmath2 which satisfy ( i ) but not ( ii ) , and are not ph distributions .
they are necessary @xmath56-rational sequences , but are not @xmath8-rational sequences as defined above , even though they are sequences of elements of @xmath8 ( probabilities ) . in abstract - algebraic terms
, this situation is possible because the semiring @xmath56 is not a _ fatou extension _ of the semiring @xmath8 @xcite .
the existence of pathological examples of this type does not vitiate the usefulness of ph distributions in run - length modeling .
most work on rna secondary structure prediction that draws on formal language theory has employed scfgs @xcite .
a cfg in chomsky normal form ( cnf ) , used for generating strings in @xmath26 where @xmath25 is a finite alphabet set , is a set of production rules of the form @xmath78 or @xmath79 , where @xmath80 are elements of a set @xmath81 of ` variables ' , i.e. , nonterminal symbols , and @xmath82 .
there is a distinguished start symbol @xmath83 with which the process begins . applying the production rules
repeatedly yields a subset @xmath84 , i.e. , a language .
an scfg assigns probabilities ( which add to unity ) to the productions of each @xmath85 , and yields a probability distribution over the strings in @xmath84 , i.e. , over @xmath26 .
the probability distribution @xmath86\subset\mathbb{r}_+$ ] produced by an scfg is an example of an @xmath8-algebraic series . in general ,
if @xmath48 is a semiring , an @xmath48-algebraic series ( of cnf type ) over an alphabet @xmath25 is a weighting function @xmath87 obtained as one component ( i.e. , the component @xmath88 ) of the formal solution of a coupled set of quadratic equations @xmath89 computed by iteration @xcite .
the coefficients @xmath90 and @xmath91 are elements of @xmath48 , so each @xmath92 is a sum of @xmath48-weighted strings in @xmath26 , or equivalently a function @xmath93 .
it is clear that theorem [ thm : normalization ] has an analogue : any probability distribution on @xmath26 produced by a scfg of cnf type is simply a _ normalized _
@xmath8-algebraic series .
any scfg of cnf type has @xmath94 parameters , which may be too many for practical estimation if a small sequence family is being modeled . to facilitate modeling
, one should use an scfg with a restricted structure , and also exploit results from weighted automata theory .
if the nucleotide distribution does not vary much along typical sequences , then the alphabet set @xmath25 can be taken to be a @xmath5-letter alphabet @xmath17 ( if one is modeling watson crick pairing , exclusively ) or even a @xmath0-letter alphabet ( if one is modeling runs of unpaired bases ) . also , one can leverage the fact that @xmath48-algebraic series subsume @xmath48-rational series , which implies ( in the @xmath0-letter case ) that @xmath48-algebraic _ sequences _ , which are effectively indexed by @xmath2 , subsume @xmath48-rational sequences . in the boolean ( @xmath66 )
case , the first statement is the familiar chomsky hierarchy . in the case of a @xmath0-letter alphabet @xmath95 , an scfg defines a probability distribution on @xmath96 .
that is , it defines an @xmath2-valued random variable @xmath12 , the length of the string emitted by the stochastic push - down automaton ( spda ) corresponding to the scfg .
the spda uses @xmath81 , the set of nonterminal symbols , as its stack alphabet , and its stack is initially occupied by the start symbol @xmath21 .
the stochastic production rules specify what happens when a symbol @xmath85 is popped off the stack : either two symbols @xmath97 are pushed back , or a letter ` @xmath19 ' is emitted . by construction , at least one letter must be emitted by a cnf - type scfg before its stack empties , so @xmath98 . the class of probability distributions on @xmath2 associated to scfgs ( whether or not of cnf type ) , i.e. , that of normalized @xmath8-algebraic sequences , is potentially useful in parametric stochastic modeling , but has not been widely employed .
it will be denoted @xmath99 here , since each distribution in it has an algebraic @xmath9-transform @xmath100 . for any scfg ,
an algebraic equation satisfied by @xmath6 can be computed by polynomial elimination ( e.g. , by computing the resultant of the above system of quadratic equations ) .
let @xmath101 denote the class of discrete phase - type distributions .
[ thm : hadamardetc ] ( i ) @xmath102 .
( ii ) if @xmath103 are independent @xmath2-valued random variables ( rvs ) with distributions in @xmath101 , then conditioning on @xmath104 yields an rv with distribution in @xmath99 .
( iii ) if , furthermore , @xmath105 is an independent @xmath2-valued rv with distribution in @xmath99 , then conditioning on @xmath106 yields an rv with distribution in @xmath99 . these are ` normalized ' versions of standard facts on @xmath48-rational and @xmath48-algebraic series , in particular on their composition under the hadamard product @xmath107 , in the special case when @xmath108 and @xmath95 .
( see @xcite . )
they have direct probabilistic proofs .
e.g. , to prove ( i ) , one would show that starting from the distribution of @xmath38 , the absorption time in a hmm , one can construct an scfg that yields the same distribution on @xmath2 .
( if @xmath109 then the scfg can not be of cnf type . )
the procedure is similar to constructing a pda that accepts a specified regular language .
much as with discrete ph distributions , it is difficult to parametrize distributions in the class @xmath99 without , rather explicitly , parametrizing the stochastic automata ( scfgs or spdas ) that give rise to them ; or at least their @xmath9-transforms .
it is difficult , in general , to characterize when a probability distribution on @xmath2 that has an algebraic @xmath9-transform lies in @xmath99 .
the following example illustrates the problem .
any distribution @xmath16 on @xmath2 that has an algebraic @xmath9-transform necessarily satisfies a finite - depth recurrence of the form @xmath110 , where the functions @xmath111 , @xmath112 are polynomial in @xmath113 .
( if none of the @xmath111 depends on @xmath113 , then the @xmath9-transform will be rational . )
consider , for example , the @xmath5-term recurrence @xmath114 where @xmath115 are parameters , which is of this form .
the @xmath9-transform @xmath100 of its solution is proportional , by definition , to @xmath116 , which is gauss s parametrized hypergeometric function .
the set of triples @xmath117 that yields an _ algebraic _ @xmath9-transform , and hence an @xmath56-algebraic sequence @xmath118 , is explicitly known .
it was derived in the nineteenth century by h. a. schwartz ( * ? ? ?
unfortunately , it is an _
infinite discrete _ subset of @xmath119 , not a continuous subset . in general
, the @xmath9-transform of the solution of a finite - depth recurrence of the above form will be algebraic in @xmath9 only if the overall parameter vector of its coefficients , the polynomials @xmath120 , is confined to a submanifold of positive codimension . for distributions in @xmath99 , this makes recurrence - based parametrization less useful than scfg - based or @xmath9-transform - based parametrization .
a new scheme for modeling the prior distribution of secondary structures in an rna sequence family will now be proposed .
it will exploit the insights of sections [ sec : duration ] and [ sec : algebraic ] , on the class of discrete phase - type distributions on @xmath2 ( i.e. , @xmath101 ) , and the larger class of @xmath8-algebraic distributions on @xmath2 ( i.e. , @xmath99 ) .
if @xmath37 is the alphabet set , any scfg , or its associated spda , will define a probability distribution on @xmath26 , the set of finite length sequences @xcite .
( the distribution of the sequence length , which is a random variable , lies in @xmath121 . ) but even if the scfg is in chomsky normal form ( cnf ) , the number of its parameters grows cubically in the number of grammar variables , as mentioned above . to facilitate estimation , the model should have a restricted structure . the models of knudsen and hein @xcite and nebel @xcite are representative .
the knudsen
hein scfg has variable set @xmath122 , and production rules @xmath123 here @xmath124 signifies an unpaired base and @xmath125 signifies two bases that are paired in the secondary structure , so @xmath126 produces runs of unpaired bases , i.e. , loops ( which may include stems ) and @xmath127 produces runs of paired bases , i.e. , stems ( which may include loops of length at least @xmath5 ) .
this scfg is not a cnf one , but model parameters may be estimated by a variant of the inside
outside algorithm .
if one takes single base frequencies and pair frequencies ( i.e. , the probability of @xmath125 representing @xmath48@xmath128 , @xmath129@xmath130 , or even @xmath129@xmath128 ) into account , one has only three independent parameters to be estimated , one probability per production rule .
knudsen and hein ( cf .
nebel ) used as their primary training set a subset of the european database of long subunit ribosomal rnas ( lsu rrnas ) @xcite . for the probabilities of @xmath131 vs. @xmath132 ( from @xmath21 ) , they estimated @xmath133 vs. @xmath134 ; for @xmath124 vs. @xmath135 ( from @xmath132 ) , @xmath136 vs. @xmath137 ; and for @xmath138 vs. @xmath131 ( from @xmath127 ) , @xmath139 vs. @xmath140 .
their training set actually included trnas as well , since they were attempting to model the family of folded rna molecules as a whole . as knudsen and hein note , their model yields loops and stems with geometrically distributed lengths . to improve quantitative agreement , it would need to be made more sophisticated .
it would also benefit from a cleaner separation between its two levels : the paired - base and unpaired - base levels , i.e. , the context - free and regular levels ( in the formal language sense ) , i.e. , the spda and hmm levels ( in the stochastic automata - theoretic sense ) . the above production rules couple the two levels together .
it is not clear from ref .
@xcite how well the model stochastically fits the length of ( i ) training sequences , ( ii ) the subsequences comprising paired bases , and ( iii ) the subsequences comprising unpaired bases . separating the two levels
should facilitate the separate fitting of these quantities . by definition , folded rna secondary structure is characterized by a subsequence comprising paired bases , so the stochastic modeling of secondary structure in a given family should initially focus on such subsequences .
if pseudo - knots ( a thorny problem for automata - theoretic modeling ) are ignored , these subsequences are effectively _ dyck words _ , or balanced parenthesis expressions . in the absence of covariation ,
one expects to be able to generate such words over @xmath141 from classical dyck words over the @xmath5-letter alphabet @xmath17 , consisting of opening and closing parentheses , by replacing each @xmath19@xmath20 pair independently by an @xmath48@xmath128 , @xmath129@xmath130 , or @xmath129@xmath128 pair , according to observed pair frequencies .
knudsen and hein note that order matters : in @xmath129@xmath130/@xmath130@xmath129 pairs in trna , the @xmath129 tends to be nearer the @xmath142 end of the rna than the @xmath130 . still to be resolved , of course , is the selection of the underlying probability distribution over dyck words in @xmath18 .
one could start with any cfg that unambiguously generates the dyck words over @xmath17 , and make it stochastic by weighting its productions .
the simplest such cfg is a @xmath0-variable one , @xmath143 .
the corresponding scfg is @xmath144 where @xmath145 .
this scfg , with @xmath146 free parameters , is so simple that it can be studied analytically .
the length of a dyck word is an @xmath2-valued random variable , the distribution of which lies in @xmath99 , with a parameter - dependent , algebraic @xmath9-transform . as was explained in section [ sec : algebraic ]
, it is best to parametrize distributions in @xmath99 by the scfgs that give rise to them , rather than by explicit formulas or even by the recurrence relations that they satisfy ; and this is an example .
this dyck model could be made arbitrarily more versatile , since arbitrarily complicated cfgs that generate the dyck language over @xmath17 can readily be constructed .
one could , for instance , iterate @xmath147 once , obtaining a production rule for @xmath21 with @xmath148 alternatives on its right - hand side . weighting them with probabilities would yield an scfg with @xmath149 independent parameters , which would be capable of much more accurate fitting of data on an empirical family of rna sequences . in general
, one could choose model parameters to fit not only the observed distribution of dyck word lengths ( i.e. , per - family paired - base subsequence lengths ) , but also the distribution of lengths of stems , i.e. , runs of contiguous paired bases , which may be far from geometric .
the preceding discussion of dyck words formed from paired bases ignored loops , i.e. , runs of unpaired bases .
they are best handled on a second level of the scfg .
the simplest production rule for _ full _ sequences would have not @xmath21 , @xmath29 , @xmath30 , @xmath31 on its right - hand side , but rather @xmath32 , @xmath33 , @xmath34 , @xmath35 , where each of the eight @xmath36s expands to a run of unpaired bases . in the absence of covariation , modeling each run is a matter of duration modeling . starting from an @xmath2-valued random run length , or equivalently a distribution over finite @xmath0-letter words ,
one would generate a run of unpaired bases by replacing each letter independently by @xmath150 , according to family - specific single base frequencies .
each run length is naturally taken to have a distribution in @xmath101 , since that will allow the resulting run of bases to be generated by an hmm ( with absorption )
. each run length will be the absorption time in a finite - state markov chain , the parameters of which , i.e. , transition probabilities , can be estimated from empirical data .
geometric distributions , and generalizations , are appropriate .
it follows from theorem [ thm:3 ] that employing a large markov chain with a fully connected transition graph , and hence a number of parameters that grows quadratically in the number of states , would _ not _ be appropriate . without loss of generality
, each transition graph can be assumed to have no ` cycles within cycles within cycles ' . in this
extended ( two - level ) stochastic model of the secondary structure of a family of rna sequences , the sequence length distribution still lies in @xmath99 .
that is because a ( dyck - type ) spda wrapped around one or more hmms is still an spda , with an scfg representation .
this observation is similar to the proof of theorem [ thm : hadamardetc ] : what has been constructed here is simply an scfg with a special structure , not given explicitly in chomsky normal form .
the full set of model parameters could be estimated by the inside
outside algorithm @xcite , rather than by estimating dyck - scfg and run - length parameters separately ; but that would not be so efficient
. the test of the proposed scfg architecture will be its value in secondary structure prediction , since from any rna sequence the most likely parse tree , and paired - base subsequence , can be computed by maximum a posteriori estimation .
t. katayama , m. okamoto , and h. enomoto , `` characterization of the structure - generating functions of regular sets and the d0l growth functions , '' _ inform . and control _ , vol .
36 , no . 1
, pp . 85101 , 1978 .
y. sakakibara , m. brown , r. hughey , i. s. mian , k. sjlander , r. c. underwood , and d. haussler , `` stochastic context - free grammars for trna modeling , '' _ nucleic acids research _ , vol .
22 , no .
51125120 , 1994 . | we propose a two - level stochastic context - free grammar ( scfg ) architecture for parametrized stochastic modeling of a family of rna sequences , including their secondary structure .
a stochastic model of this type can be used for maximum a posteriori estimation of the secondary structure of any new sequence in the family .
the proposed scfg architecture models rna subsequences comprising paired bases as stochastically weighted dyck - language words , i.e. , as weighted balanced - parenthesis expressions .
the length of each run of unpaired bases , forming a loop or a bulge , is taken to have a phase - type distribution : that of the hitting time in a finite - state markov chain . without loss of generality , each such markov chain can be taken to have a bounded complexity .
the scheme yields an overall family scfg with a manageable number of parameters . |
understanding the origin of life is one of the main challenges of modern science .
it is believed that some basic prebiotic chemistry could have developed in space , likely tranferring prebiotic molecules to the solar nebula , which were finally delivered to earth .
studies of the chemical composition of comets have indeed reported that these objects exhibit a rich chemistry in complex organic molecules ( or coms ) that are commonly detected in the ism ( see e.g. , * ? ? ?
recently , the spacecraft rosetta studied the chemical composition of the comet 67p / churyumov - gerasimenko , and found 16 different coms of prebiotic interest such as glycoladehyde ( ch@xmath6ohcho ) , formamide ( nh@xmath6cho ) @xcite , and even the simplest amino acid glycine @xcite .
interestingly , a new molecule was detected by the rosetta mission with a relatively high abundance compared to other coms present in the comet : the simplest isocyanate ( methyl isocyanate , ch@xmath0nco ; * ? ? ?
* ) , which like nh@xmath6cho ( the simplest amide ) contains c , n , and o atoms .
isocyanates are a family of prebiotic coms that could play an key role in the synthesis of aminoacid chains known as peptides @xcite .
subsequent to the detection of ch@xmath0nco in comet 67p / churyumov - gerasimenko , this molecule was also detected in the massive hot molecular cores sgrb2
n @xcite and orion kl @xcite .
however , ch@xmath0nco has not been detected so far toward solar - type protostars , the environment where we expect earth - like planets to form .
iras 16293@xmath82422 ( hereafter iras16293 ) is located in the @xmath9 ophiuchi cloud complex at a distance of 120 pc @xcite , and it is considered an excellent low - mass protostar template for astrochemical studies ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
its molecular spectra is rich in coms which has allowed the detection of molecular species such as glycoladehyde and formamide ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , along with deuterated molecules @xcite , and exotic species such as ch@xmath0sh @xcite .
iras16293 is a binary system with components a and b separated in the plane of the sky by 5@xmath10 ( @xmath7600 au ) , and whose masses are @xmath70.5 m@xmath11 @xcite .
their emission exhibits line profiles with different linewidths of up to 8 km s@xmath12 for iras 16293 a and @xmath13 2 km s@xmath12 for iras 16293 b. the narrow emission of iras 16293 b , along with its rich com chemistry , makes this object the perfect target to search for new coms . in this letter , we report the first detection of ch@xmath0nco towards the solar - type protostar iras 16293@xmath82422 b carried out with the atacama large millimeter array ( alma ) .
the comparison between the derived column density and our chemical modelling suggests that ch@xmath0nco can be formed via gas - phase reactions after the evaporation of hnco from dust grains . [
cols="^,^,^,^,^,^,^,^ " , ] in this model , hnco is formed mainly via hydrogenation on the surface of dust grains , and once the temperature of the dust reaches a temperature of @xmath7100 k , hnco is thermally desorbed and incorporated into the gas phase
. the gas - phase chemistry formation of ch@xmath0nco can thus proceed ( see reactions above and * ? ? ?
* ) , yielding a ch@xmath0nco abundance very close to the value measured toward iras16293 b ( of @xmath14 ) in this study . since no other mechanisms are required to explain the observed abundance of this molecule , we propose that gas phase formation may be the dominant mechanism for the production of ch@xmath0nco . we also stress that the predicted abundance of ch@xmath0nco does not depend on the maximum temperature assumed in our model ( we obtain the same results for t = 100 k and t = 250 k ) , as long as this temperature is high enough ( @xmath15100 k)to induce the full evaporation of the ices and the conversion of hnco into ch@xmath0nco
alternatively , ch@xmath0nco could be produced by uv photoprocessing of hnco - containing interstellar ices , and subsequently be desorbed to the gas phase @xcite .
this scenario is not taken into account in our chemical model , but could be needed if any ch@xmath0nco gas - phase destruction route were included , thus requiring an additional source of ch@xmath0nco molecules .
we present the first detection of ch@xmath0nco toward a solar - type protostar .
the derived abundance of 7@xmath210@xmath5 is similar to that measured in hot cores , although the column density ratios with other chemically related n- , c- , o - bearing species do not show an agreement overall .
the ch@xmath0nco / nh@xmath6cho column density ratio is the only one that agrees best across all environments in which ch@xmath0nco has been detected so far ( hot corinos , hot cores and comet 67p / churyumov - gerasimenko ) . the observed ch@xmath0nco abundance in iras16293 b is well reproduced by a chemical model taking into account only the gas - phase formation of this species , although an origin in the solid phase remains to be explored .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) , nsc and asiaa ( taiwan ) , and kasi ( republic of korea ) , in co - operation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj this research was partially financed by the spanish mineco under project aya2014 - 60585-p , by the italian ministero dell istruzione , universit e ricerca , through the grant progetti premiali 2012 ialma ( cup c52i13000140001 ) , and by the gothenburg centre for advanced studies in science and technology , where the re - calibration and re - imaging of all the alma archive data on iras16293 - 2422 was carried out as part of the gocas program `` origins of habitable planets '' .
d . benefited from a fpi grant from spanish mineco .
i.j .- s . acknowledges the financial support received from the stfc through an ernest rutherford fellowship ( proposal number st / l004801/1 ) .
p . acknowledges partial support by the mineco under grants fis2012 - 39162-c06 - 01 , esp2013 - 47809-c03 - 01 , and esp2015 - 65597-c4 - 1 .
altwegg , k. , balsiger , h. , bar - nun , a. , et al .
2015 , science , 347 , 6220 , id.1261952 + belloche , a. , menten , k.m . , comito , c. , et al .
2008 , a&a , 482 , 1 , 179 + beltrn , m. t. , codella , c. ; viti , s. , neri , r. , & cesaroni , r. 2009 , apj , 690 , 2 , l93 + bisschop , s.e . ,
jorgensen , j.k . , bourke , t.l .
, bottinelli , s. , & van dishoeck , e.f .
2008 , a&a , 488 , 3 , 959 + biver , n. , bockele - morvan , d. , debout , v. , et al .
2014 , a&a , 516 , a109 + bottinelli , s. , ceccarelli , c. , neri , r. , et al .
2004 , apj , 617 , 1 , l69 + caux , e. , kahane , c. , castets , a. , et al .
2011 , a&a , 532 , a23 + cazaux , s. , tielens , a.g.g.m . , ceccarelli , c. , et al .
2003 , apj , 593 , 1 , l51 + cernicharo , j. , kisiel , z. , tercero , b. , et al .
2016 , a&a , 587 , l4 + coutens , a. , jorgensen , j. , van der wiel , m.h.d . ,
et al . 2016 , a&a , 590 , l6 + coutens , a. , vastel , c. , caux , e. , et al .
2012 , a&a , 539 , a132 + goesmann , f. , rosenbauer , h. , bredehft , j.h . ,
2015 , science , 346 , 6247 , id.0689 + halfen , d.t .
, ilyushin , v.v . , & ziurys , l.m .
2015 , apjl , 8212 , l5 + hasegawa , t. i. , herbst , e. , & leung , c. m. 1992 , , 82 , 167 hollis , j. m. , lovas , f. j. , & jewell , p. r. 2000 , apj , 540 , 2 , l107 + jorgensen , j.k . , bourke , t.l . , nguyen luong , q. , & takakuwa , s. 2011 , a&a , 534 , a100 + jorgensen , j.k . , favre , c. , bisschop , s.e . ,
bourke , t.l .
, et al . 2012 , apjl , 757 , 1 , l4 + jorgensen , j.k . , van der wiel , m.h.d . , coutens , a. , et al .
2016 , a&a , 595 , a117 + kahane , c. , ceccarelli , c. , faure , a. , & caux , e. 2013 , apjl , 763 , 2 , l38 + kasten , w. , & dreizler , h. 1986 , z. natur . 41a , 637 + koput , j. 1986 , jmosp , 115 , 131 + kuan , y .- j . , huang , h .- c . ,
charnley , s.b . , et al .
2004 , 616 , 1 , l27 + loinard , l. , torres , r.m . ,
mioduszewski , a.j . , &
rodrguez , l.f .
2008 , apjl , 675 , 1 , l29 + looney , l.w . ,
mundy , l.g . , & welch , w.j .
2000 , apj , 529 , 1 , 477 + lpez - sepulcre , a. , jaber , a.a . ,
mendoza , e. , et al .
2015 , mnras , 449 , 3 , 2438 + lykke , j.m .
, coutens , a. , jorgensen , j. , et al .
2016 , in press .
rivilla , v.m . ,
fontani , f. , beltrn , m.t . , et al .
2016 , apj , 826 , 2 + rivilla , v. m. and beltrn , m. t. , cesaroni , r. , et al . , 2016 ,
a&a , in print , arxiv:1608.07491 + roberts , j. f. , rawlings , j. m. c. , viti , s. , & williams , d. a. 2007 , , 382 , 733 | we report the first detection of the prebiotic complex organic molecule ch@xmath0nco in a solar - type protostar , iras16293 - 2422 b. this species is one of the most abundant complex organic molecule detected on the surface of the comet 67p / churyumov - gerasimenko , and in the insterstellar medium it has only been found in hot cores around high - mass protostars .
we have used multi - frequency alma observations from 90 ghz to 350 ghz covering 11 unblended transitions of ch@xmath0nco and 40 more transitions that appear blended with emission from other molecular species .
our local thermodynamic equilibrium analysis provides an excitation temperature of 232@xmath141 k and a column density of ( [email protected])@xmath210@xmath3 @xmath4 , which implies an abundance of ( 7@xmath12)@xmath210@xmath5 with respect to molecular hydrogen .
the derived column density ratios ch@xmath0nco / hnco , ch@xmath0nco / nh@xmath6cho , and ch@xmath0nco / ch@xmath0cn are @xmath70.3 , @xmath70.8 , and @xmath70.2 , respectively , which are different from those measured in hot cores and in comet 67p / churyumov - gerasimenko .
our chemical modelling of ch@xmath0nco reproduces well the abundances and column density ratios ch@xmath0nco / hnco and ch@xmath0nco / nh@xmath6cho measured in iras16293 - 2422 b , suggesting that the production of ch@xmath0nco could occur mostly via gas - phase chemistry after the evaporation of hnco from dust grains . |
the quark model predicts the existence of heavy baryons composed of single , double and triple quarks .
essential improvement has been achieved in heavy baryon spectroscopy in resent years .
all baryons with a single charm quark that are predicted by the quark model have been observed in experiments .
moreover , heavy baryons with a single bottom quark , such as @xmath2 , @xmath3 , @xmath4 and @xmath5 have also been discovered ( for a review see @xcite ) . these progresses in experiments stimulated future investigation for the properties of these baryons at lhc , as well as further theoretical studies on this subject .
remarkable information about the internal structure of baryons can be gained by studying their magnetic moments .
magnetic moments of the heavy baryons have long been under the focus of theoretical physicists , and they have been studied in framework of various approaches .
so far , magnetic moments of charmed heavy baryons have been calculated in framework of the naive quark model @xcite , relativistic quark model @xcite , chiral perturbation model @xcite , hypercentral model @xcite , soliton model @xcite , skyrmion model @xcite , bag model @xcite , qcd sum rules method @xcite , nonrelativistic quark model @xcite , phenomenological relativistic model @xcite , quark model with confinement law potential @xcite , and chiral constituent quark model , respectively .
magnetic moments of heavy hadrons have also been studied in numerous works within the qcd sum rules method .
magnetic moments of the @xmath6 and @xmath7 baryons have been calculated in framework of the traditional qcd sum rules method in @xcite ; of the @xmath8 and @xmath9 @xmath10 or @xmath11 baryons within the light version of the qcd sum rules method in @xcite , respectively .
it should be noted here that magnetic moments of the spin-3/2 heavy baryons have already been analyzed within the same approach in @xcite .
the goal of the present work is calculation of the magnetic moments of heavy @xmath12 , @xmath13 and @xmath14 baryons within the light cone qcd sum rules method .
the paper is organized as follows . in section 2 ,
sum rules for the magnetic moments of the above - mentioned baryons are constructed .
numerical results of our calculations are presented in section 3 .
this section further contains comparison of our results with the predictions of other approaches , and concluding remarks .
we start this section by giving a brief summarizing of the classification of heavy baryons in su(3 ) symmetry . according to su(3 ) symmetry ,
heavy baryons with single heavy quark belong to either symmetric sextet or antisymmetric anti - triplet flavor representations .
as has already been noted , magnetic moments of the @xmath8 and @xmath9 baryons belonging to anti - symmetric flavor representations of su(3 ) are calculated within the light qcd sum rules method in @xcite and @xcite , respectively . in the present work we calculate magnetic moments of the heavy baryons belonging to sextet representation of su(3 ) group . in order to calculate magnetic moments of the heavy sextet baryons ,
we start by considering the following correlation function : [ enil01 ] ( p , q ) = i d^4x e^ipx ( q ) \ { _ q ( x ) |_q(0 ) } 0 , where @xmath15 is the interpolating current of the heavy spin-1/2 baryon .
the main task in constructing the sum rules for magnetic moments of the heavy sextet baryons is the calculation of the correlation function in terms of the photon distribution amplitude by using the operator product expansion ( ope ) over twist from one side , and in terms of the hadrons from the other side , and then equating both representations .
calculation of the correlation function from the hadronic side is accomplished by inserting a complete set of hadrons carrying the same quantum numbers of the interpolating current @xmath15 . isolating the the ground state
s contribution , we obtain , [ enil02 ] = 0 _ q b_q ( p_2 ) p_2 ^ 2-m_b_q^2 b(p_2 ) ( q ) b_q ( p_1 ) _ + , where dots correspond to the contributions of higher states and continuum , @xmath16 . in further analysis we shall make the replacement @xmath17 .
the matrix elements in eq .
( [ enil02 ] ) are determined as , [ enil03 ] 0 _ q b_q ( p ) _ q
u_b_q(p ) , + [ enil04 ] b_q(p ) b_q ( p_1 ) _ ^|u_b_q ( p ) u_b_q(p_1 ) , where @xmath18 is the photon polarization vector , @xmath19 and @xmath20 are the form factors . using the equation of motion , eq .
( [ enil04 ] ) can be written as , [ enil05 ] b_q(p ) b_q ( p_1 ) _ = |u_b_q ( p ) u_b_q ( p_1 ) ^ .
since the photon involved in these transitions is a real photon , we set @xmath21 , and hence in calculation of the magnetic moments of the heavy sextet baryons the values of the form factors are needed only at @xmath21 . substituting eqs .
( [ enil03 ] ) and ( [ enil05 ] ) in eq .
( [ enil02 ] ) , and performing summations over spins of the heavy baryons , for the hadronic of the correlation function we get , [ enil06 ] = _ b_q^2 ^ .
we observe from eq .
( [ enil06 ] ) that the correlation function contains many structures , any of them can be chosen in calculating magnetic moments of the sextet baryons , and in this respect we choose the structure @xmath22 .
this structure envelopes the magnetic form factor @xmath23 , and at @xmath21 it gives the magnetic moment @xmath24 of the heavy baryons in units of @xmath25 . as a result , the correlation function can be written in terms of the magnetic moment of heavy baryons as , [ enil07 ] = _
b_q^2 1m_b_q^2 - p^2 _ b_q 1m_b_q^2 - ( p+q)^2 . in order to calculate the correlation function in terms of quark and gluon degrees of freedom and photon distribution amplitudes ,
the expressions of the interpolating currents of the heavy baryons are needed .
the general form of the interpolating currents of the heavy spin-1/2 positive parity baryons is given as ( see for example @xcite ) , [ enil08 ] _ b_q - 1 ^abc \ { ( q_1^at c q^b ) _ 5 q_2^c + t ( q_1^at c _ 5 q^b ) q_2^c - ( q^at c q_2^b ) _ 5 q_1^c - t ( q^at c _ 5 q_2^b ) q_1^c } , where @xmath26 are the color indices
, @xmath27 is the charge conjugation operator , and @xmath28 is an arbitrary parameter whose value at @xmath29 gives the so - called ioffe current .
the quark contents of the sextet heavy sextet baryons are given in table 1 .
@xmath30 using the expression for the interpolating current and wick s theorem , the theoretical part of the correlation function for the @xmath31 can be written as , [ enil09 ] ^_b^0 -3 \ { _ 5 s_d(x )
s_b^(x ) s_u(x ) _ 5 + _ 5 s_u(x ) s_b^(x ) s_d(x ) _ 5 + _ 5 s_u(x ) _ 5 tr + _ 5 s_d(x ) _ 5 tr + t ( _ 5 s_d(x ) _ 5 s_b^(x ) s_u(x ) + _ 5 s_u(x ) _ 5 s_b^(x ) s_d(x ) + s_d(x ) s_b^(x ) _ 5 s_u(x ) _ 5 + s_u(x )
s_b^(x ) _ 5 s_d(x ) _ 5 + _ 5
s_u(x ) tr + s_u(x ) _ 5 tr + _ 5 s_d(x ) tr + s_d(x ) _ 5 tr ) + t^2 ( s_d(x ) _ 5 s_b^(x )
_ 5 s_u(x ) + s_u(x ) _ 5 s_b^(x ) _ 5 s_d(x )
+ s_d(x ) tr + s_u(x ) tr ) } , where @xmath32 , @xmath33 symbolizes transposition operator , and @xmath34 is the quark ( light or heavy ) propagator .
the corresponding expressions of the correlation functions for the other members of the sextet baryons can be found from the @xmath35 by making the following replacements : [ nolabel95 ] ^_c^0 ^_c^+(u d ) , + ^_c^++ ^_c^+(d u ) , + ^_b^- ^_b^0(u d ) , + ^_b^+ ^_b^0(d u ) , + ^_c^+ ^_c^+(d s ) , + ^_c^0 ^_c^+(u d , d s ) , + ^_b^0 ^_b^0(d s ) , + ^_b^- ^_b^0(u d , d s ) , + ^_c^0 ^_c^+(u s , d s ) , + ^_b^- ^_b^0(u s , d s ) . the correlation function described by eq .
( [ enil09 ] ) contains three different parts : a ) perturbative part , when a photon is radiated from short distance ( perturbative contribution ) ; b ) a photon is radiated from short distance from the quark propagators , and light quarks form a condensate ; c ) a photon is radiated from long distance ( nonperturbative contribution ) . in order to calculate the perturbative contribution of the correlation function ,
it is enough to make the replacement [ nolabel01 ] s d^4y s^free(x - y ) ( y ) s^free(y ) , for one of the quark propagator , where @xmath36 is the free quark operator ; and the other two are being the free propagators .
the expressions of the free light and heavy quarks in coordinate representation are given as , [ nolabel02 ] s_q^free - m_q 4 ^2 x^2 , + s_q^free + i m_b_q^2 4 ^2 k_2 ( m_b_q ) ( ) ^2 , where @xmath37 are the bessel functions . the contribution of part ( c ) can easily be calculated by replacing one of the light quark operator with , [ nolabel03 ] ( s_q^ab ) _ - 14 ^ab |q^a _
i q^b ( _ i ) _ , where @xmath38 are the full set of dirac matrices , and other quark operators that involve perturbative , as well as nonperturbative terms . the explicit forms of the full " quark propagators can be found in @xcite and @xcite .
nonperturbative contribution is realized as the matrix element of the nonlocal operator @xmath39 between the vacuum and one - photon states.these matrix elements are described in terms of photon distribution amplitudes , as are given below ( see @xcite ) : [ nolabel04 ] & & ( q ) |q(x ) _
q(0 ) 0 = -i e_q |q q ( _ q_- _ q _ ) _ 0 ^ 1 du e^i |u qx ( _ ( u ) + ( u ) ) + & & - e_q _ 0 ^ 1 du e^i |u
q x h_(u ) + & & ( q ) |q(x ) _
q(0 ) 0 = e_q f_3 ( _ - q _ ) _ 0 ^ 1 du e^i |u q x ^v(u ) + & & ( q ) |q(x ) _ _ 5 q(0 ) 0 = - e_q f_3 _ ^q^x^_0 ^ 1 du e^i |u q x ^a(u ) + & & ( q ) | |q(x ) g_s g _ ( v x ) q(0 ) 0 = -i e_q ( _ q_- _ q _ ) _ i e^i ( _ |q + v _ g ) q x s(_i ) + & & ( q ) | |q(x ) g_s g _ i _ 5 ( v x ) q(0 ) 0 = -i e_q ( _ q_- _ q _ ) _ i e^i ( _ |q + v _ g ) q x ( _ i ) + & & ( q ) |q(x ) g_s g_(v x ) _ _ 5 q(0 ) 0 = e_q f_3 q_(_q_- _ q _ ) _ i e^i ( _ |q + v _ g ) q x a(_i ) + & & ( q ) |q(x ) g_s g_(v x ) i _ q(0 ) 0 = e_q f_3 q_(_q_- _ q _ ) _ i e^i ( _ |q + v _
g ) q x v(_i ) + & & ( q ) |q(x ) _
g_s g_(v x ) q(0 ) 0 = e_q \ { _ i e^i ( _ |q + v _
g ) qx t_1(_i ) + & & + _
i e^i ( _ |q + v _
g ) qx t_2(_i ) + & & + ( q_x_- q_x _ ) ( _ q_- _ q _ ) _ i e^i ( _ |q + v _
g ) qx t_3(_i ) + & & + .
( q_x_- q_x _ ) ( _ q_- _ q _ ) _ i e^i ( _ |q + v _ g ) qx t_4(_i ) } . in the definitions given above
, @xmath40 is the magnetic susceptibility of the quarks , @xmath41 is the leading twist-2 , @xmath42 , @xmath43 , @xmath44 and @xmath45 are the twist-3 , and @xmath46 , @xmath47 , @xmath48 ( @xmath49 ) are the twist-4 photon das , respectively , whose explicit expressions are given in appendix a. the measure @xmath50 is defined as [ nolabel05 ] _ i = _ 0 ^ 1 d _ |q _ 0 ^ 1 d _ q _ 0 ^ 1 d _ g ( 1-_|q-_q-_g ) . in constructing the sum rules for the magnetic moments of heavy sextet baryons it is necessary to equate the coefficients of the structure @xmath22 to from the phenomenological and theoretical representations of the correlation function .
the following steps in obtaining the final form of the sum rules for the magnetic moment are : fourier transformation , borel transformation over the variables @xmath51 and @xmath52 variables , and continuum subtraction in order to suppress the contribution of the higher states and continuum . after these procedures we get the sum rules for the magnetic moment of the sextet heavy baryons , which can schematically be written in the following form , [ nolabel06 ] _
b_q_i e^m_b_q_i^2/m^2 = ^b , theor , where the function @xmath53 contain perturbative and nonperturbative contributions .
the expression for @xmath53 is quite lengthy , and for this reason we do not present it here . in order to determine the magnetic moments of heavy sextet baryons ,
the value of the overlap amplitude is needed , which can be found from the analysis of the two - point correlation function .
the residues of the heavy sextet baryons are calculated in @xcite , and their expressions can be found in this work .
having obtained the sum rules for the magnetic moments of the heavy sextet spin-1/2 baryons , we are now ready to perform the numerical analysis .
the main input parameters of the light cone qcd sum rules for the magnetic moments are the photon distribution functions das @xcite .
their explicit expressions are presented in appendix a. the values of the other input parameters are : @xmath54 , @xmath55 @xcite , @xmath56 @xcite , @xmath57 @xcite @xmath58 @xcite , @xmath59 @xcite .
few words about the magnetic susceptibility involved in the numerical analysis are in order .
its value is estimated to have the value @xmath60 in @xcite . in @xcite , using the vector dominance model ansatz and qcd sum rules its value is predicted to be @xmath61 .
furthermore , from an analysis of the radiative decays of heavy mesons its value is found to be @xmath62 @xcite . in numerical analysis
we have used all these predicted values of the magnetic susceptibility .
it should also be remembered that , the sum rules for the magnetic moments involve the following auxiliary parameters : the arbitrary parameter @xmath28 appearing in the expressions of the interpolating currents , borel mass parameter @xmath63 , and the continuum threshold @xmath64 .
the magnetic moment must be independent of these parameters . in order to find regions " of these parameters
where magnetic moment exhibits good stability with respect to their variations we proceed as follows .
firstly , we try to find the upper and lower bounds of borel mass parameter at fixed values of @xmath64 and @xmath28 .
the upper bound of @xmath63 can be determined by requiring that the contribution due to the continuum threshold should be less than half of the contribution coming from the perturbative part .
the lower bound of @xmath63 can be found by demanding that the higher powers of @xmath65 contribute less than the leading twist contributions .
our numerical analysis shows that these two conditions are satisfied simultaneously if @xmath63 ranges in the following regions : [ nolabel09 ] & & 2.0 gev^2 m^2 3.0 gev^2 _ c + & & 2.2 gev^2 m^2 3.4 gev^2 _
c^ + & & 2.5 gev^2 m^2 4.0 gev^2 _ c + & & 5.0 gev^2 m^2 6.0 gev^2 _ b + & & 5.0 gev^2 m^2 6.4 gev^2 _
b^ + & & 5.2 gev^2 m^2 7.0 gev^2 _ b in regard to the continuum threshold @xmath64 appearing in the sum rules , it is known that the values of this arbitrary parameter is related to the energy of first excited state . the difference @xmath66 , where @xmath67 is the ground state mass of the baryon , is the energy needed to excite the particle to its first energy state .
this quantity usually varies between @xmath68 and @xmath69 .
analysis of the mass sum rules shows that , in order to reproduce the experimental mass of the sextet baryons the continuum threshold should vary in the following domain : [ nolabel10 ] = \ { c ( 3.10.1 ) gev _ c + ( 3.20.1 ) gev _ c^ + ( 3.30.1 ) gev _ c + ( 6.60.2 ) gev _ b + ( 6.70.2 ) gev _ b^ + ( 6.80.2 ) gev _ b .
having determined the sum rules , input parameters , and working regions of all auxiliary parameters , we perform numerical analysis to calculate the magnetic moment @xmath24 of the heavy sextet spin-1/2 baryons , whose results we can summarize as follows . as examples , in figs . 1 and 2 we present the dependence of @xmath70 on @xmath63 at several fixed values of @xmath28 and at the fixed value of @xmath64 listed above for the baryons @xmath71 and @xmath72 , respectively .
we observe from these figures that the magnetic moments of @xmath12 show good stability with respect to the variation of the borel mass parameter in aforementioned domains .
the next step in finding the values of the magnetic moments of the baryons under consideration is to determine working region of the arbitrary parameter @xmath28 , where @xmath24 is practically independent of its variation .
for this purpose , in figs . 3 and 4 the dependence of @xmath70 on @xmath73 , where @xmath74 for the @xmath71 and @xmath72 baryons , at several fixed values of the borel mass parameter @xmath63 and at the predetermined value of the continuum threshold @xmath64 , are presented , respectively .
it follows from these figures that the magnetic moments of the heavy sextet spin-1/2 baryons seem to be practically independent of the arbitrary parameter @xmath28 in the domain @xmath75 , and insensitive to the the chosen values of @xmath64 .
our final results on the magnetic moments of the heavy sextet spin-1/2 baryons are presented in table 2 . for comparison , in the same table
we present the predictions of other approaches on the magnetic moments of the relevant baryons , such as nonrelativistic quark model @xcite , bag model @xcite , phenomenological relativistic quark model @xcite , quark model with confinement law potential @xcite , relativistic quark model @xcite , hypercentral model @xcite , chiral constituent model @xcite , and qcd sum rules method @xcite .
@xmath76 we observe from this table that almost all approaches give , more or less , similar predictions , except the results of @xcite and @xcite ( especially for the charmed baryons ) , which are smaller . using our results
one can easily deduce the following relations among the magnetic moments of heavy baryons .
[ nolabel11 ] _ _ c^++ + _ _ c^0 2 _ _
c^+ , + _ _
c^++ + _ _ c^0 2 _ _
c^+ , + _
_ c^++ + 2 _ _ c^0 _ _ c^0 + 2 _ _ c^+ , + _ _
b^+ + _ _ b^- 2 _ _
b^0 , + _ _
b^+ + _ _ b^- 2 _ _ b^0 , + _ _ b^+ + 2 _ _ b^- _ _ b^- + 2 _ _
b^0 , the direct measurement of the magnetic moments of heavy baryons are unlikely in the near future .
therefore , any indirect estimations of the magnetic moments of the heavy baryons could be very useful .
for example , it could help extracting information about the mass spectrum of the heavy baryons .
as is well known , experimentally measured mass difference is attributed to the hyperfine splitting . moreover , the magnetic moments of quarks are proportional to the chromomagnetic moments , which determine the hyperfine splitting in the baryon spectrum . following this reasoning , the magnetic moments of the @xmath6 and @xmath2 baryons are estimated in @xcite . hence , this approach could be an essential tool in estimating the magnetic moments of the heavy baryons . in conclusion ,
the magnetic moments of the heavy sextet @xmath77 baryons are calculated in framework of the light cone qcd sum rules method .
empirically , linearly independent relations among the magnetic moments of the sextet baryons are obtained .
comparison of our results with the predictions of other approaches is presented .
explicit forms of the photon das @xcite . [ nolabel27 ] _ ( u ) 6 u |u , + ^v(u ) 3 [ 3 ( 2 u - 1)^2 -1 ] + ( 15 w^v_- 5 w^a _ ) [ 3 - 30 ( 2 u - 1)^2 + 35 ( 2 u -1)^4 ] , + ^a(u ) ( 1 + w^v_- w^a _ ) , + a(_i ) 360 _ q _ |q _ g^2 , + v(_i ) 540 w^v_(_q - _ |q ) _ q _ |q _ g^2 , + h_(u ) - 10 ( 1 + 2 ^+ ) c_2^(u - |u ) , + ( u ) 40 u^2 |u^2 ( 3 - ^+ + 1 ) + 8 ( _ 2^+ - 3 _ 2 ) [ u |u ( 2 + 13 u |u ) + 2 u^3 ( 10 -15 u + 6 u^2 ) ( u ) + 2 |u^3 ( 10 - 15 |u + 6 |u^2 ) ( |u ) ] , + t_1(_i ) -120 ( 3 _ 2 + _ 2^+)(_|q - _ q ) _
|q _ q _ g , + t_2(_i ) 30 _ g^2 ( _ |q - _ q ) [ ( - ^+ ) + ( _ 1 - _ 1^+)(1 - 2_g ) + _ 2 ( 3 - 4 _ g ) ] , + t_3(_i ) - 120 ( 3 _ 2 - _ 2^+)(_|q -_q ) _ |q _ q _ g , + t_4(_i ) 30 _ g^2 ( _ |q - _ q ) [ ( + ^+ ) + ( _ 1 + _ 1^+)(1 - 2_g ) + _ 2 ( 3 - 4 _ g ) ] , + s(_i ) 30_g^2\{(+ ^+)(1-_g)+(_1 + _ 1^+)(1 - _ g)(1 - 2_g ) + _ 2 [ 3 ( _ |q - _ q)^2-_g(1 - _ g ) ] } , + ( _ i ) -30_g^2\{(- ^+)(1-_g)+(_1 - _ 1^+)(1 - _ g)(1 - 2_g ) + _ 2 [ 3 ( _ |q - _ q)^2-_g(1 - _ g)]}. the parameters entering the above da s are borrowed from @xcite whose values are @xmath78 , @xmath79 , @xmath80 , @xmath81 , @xmath82 , @xmath83 , @xmath84 , @xmath85 , and @xmath86 .
99 k. olive _ et .
al _ , particle data group , chin .
c * 38 * , 090001 ( 2014 ) .
a. l. choudry , v. joshi , phys .
d * 13 * , 3115 ( 1976 ) .
d. b. lichtenberg , phys .
d * 15 * , 345 ( 1977 ) .
a. faessler _ et .
al _ , phys .
d * 73 * , 094013 ( 2006 ) .
m. savage , phys .
b * 326 * , 303 ( 1994 ) .
b. patel , a. k. rai , p.c .
vinodkumar , j. phys .
g * 35 * , 065001 ( 2008 ) .
d. o. riska , nucl .
inst . method .
b * 119 * , 259 ( 1996 ) .
y. oh , d. p. min , m. rho , n. n. scoccola , nucl . phys .
a * 534 * , 493 ( 1991 ) .
a. bernotas , and v. simonis , lithuanian journal of phys .
* 53 * , 84 ( 2013 ) .
s. l. zhu , w. y. hwang , and z. s. yang , phys .
d * 56 * , 7273 ( 1997 ) .
j. franklin , d. b. lichtenberg , w. mamgung , and d. c. carydas , phys .
d * 24 * , 2910 ( 1981 ) .
b. julia - diaz , and d. o. riska , nucl .
a * 739 * , 69 ( 2004 ) .
m. barik , and h. das , phys .
d * 29 * , 2823 ( 1983 ) .
n. sharma , p. k. chatley , and m. gupta , phys .
d * 81 * , 073001 ( 2010 ) .
t. m. aliev , a. zpineci , m. savci , phys .
d * 65 * , 096004 ( 2002 ) .
t. m. aliev , k. azizi , m. savci , phys .
d * 77 * , 114006 ( 2008 ) .
t. m. aliev , k. azizi , a. zpineci , nucl .
b * 808 * , 137 ( 2009 ) .
e. bagan , m. chabab , h. g. dosch , s. narison , phys .
b * 278 * , 369 ( 1992 ) .
i. i. balitsky , v. m. braun , nucl . phys .
b * 311 * , 541 ( 1998 ) .
v. m. braun , j. e. filyanov , z. phys .
c * 48 * , 239 ( 1990 ) .
p. ball , v. m. braun , and n. kivel , nucl .
b * 649 * , 263 ( 2003 ) .
t. m. aliev , k. azizi , a. zpineci , phys .
d * 79 * , 056005 ( 2009 ) .
v. m. belyaev , b. l. ioffe , jetp * 56 * , 493 ( 1982 ) .
k. g. chetyrkin , a. khodjamirian , a. a. pirovarov , phys .
b * 661 * , 250 ( 2008 ) . c. a. dominguez , n. f. nasrallah , r. rontisch , k. schilcher , j. high energy phys .
* 05 * 020 , ( 2008 ) .
v. m. belyaev , and y. i. kogan , yad . fiz . *
40 * , 1035 ( 1984 ) .
j. rohrwild , j. high energy phys . * 09 * , 073 ( 2007 ) .
m. karliner , h. lipkin , phys .
b * 660 * , 539 ( 2008 ) .
* fig . ( 1 ) * the dependence of the magnetic moment of the @xmath71 baryon on @xmath63 , at several fixed values of @xmath28 , and at @xmath87 .
+ + * fig .
( 2 ) * the same as fig . ( 1 ) , but for the @xmath72 baryon , at @xmath88 .
+ + * fig .
( 3 ) * the dependence of the magnetic moment of the @xmath71 baryon on @xmath73 , at several fixed values of @xmath63 , and at @xmath87 .
+ + * fig .
( 4 ) * the same as fig . ( 3 ) , but for the @xmath72 baryon , at @xmath88 . | the magnetic moments of heavy sextet @xmath0 baryons are calculated in framework of the light cone qcd sum rules method .
linearly independent relations among the magnetic moments of these baryons are obtained .
the results for the magnetic moments of heavy baryons obtained in this work are compared with the predictions of the other approaches .
# 1#2#3 @xmath1 # 1#2#3 @xmath1 0^*0 5_5 q 5_5 o _ ^0 _ pacs numbers : 11.55.hx , 13.40.em , 14.20.mr , 14.20.lq |
the most important event of last 15 years in astrophysics is conclusion about accelerated expansion of our universe at late stage of its evolution .
this conclusion was based on observations of luminosity distances and redshifts for the type ia supernovae @xcite , cosmic microwave background @xcite , large - scale galaxy clustering @xcite , and other evidence @xcite . to explain accelerated evolution of the universe various mechanisms
have been suggested , including the most popular cosmological model @xmath0cdm with a @xmath0 term ( dark energy ) and cold dark matter ( see reviews @xcite ) .
the @xmath0cdm model with 4% fraction of visible ( baryonic ) matter nowadays , 23% fraction of dark matter and 73% fraction of dark energy @xcite describes type ia supernovae , data rather well and satisfies observational evidence , connected with rotational curves of galaxies , galaxy clusters and anisotropies of cosmic microwave background .
however , the @xmath0cdm model ( along with vague nature of dark matter and energy ) has some problems with fine tuning of the observed value of @xmath0 , which is many orders of magnitude smaller than expected vacuum energy density , and with different time dependence of dark energy @xmath1 and material @xmath2 fractions ( we have @xmath3 nowadays ) . therefore a large number of alternative cosmological models have been proposed .
they include theories with extra dimensions @xcite ; matter with nontrivial equations of state , for example , chaplygin gas @xcite ; scalar fields with a potential @xcite ; modified gravity with @xmath4 lagrangian @xcite and many others @xcite . in this paper
we explore the cosmological model with anisotropic pressure and nontrivial equation of state in @xmath5 dimensions , suggested by pahwa , choudhury and seshadri in ref .
the authors omitted the important case @xmath6 , we include it into consideration .
we also analyze how to modify the equation of state and to avoid `` the end of the world '' ( the finite - time future singularity ) which is inevitable in the model @xcite . in this model
the @xmath5 dimensional spacetime is symmetric and isotropic in two subspaces : in 3 usual spatial dimensions and in @xmath7 extra dimensions .
it has the following metric with two robertson walker terms
@xcite : @xmath8 here the signature is @xmath9 , the speed of light @xmath10 , @xmath11 and @xmath12 is are the scale factor and curvature sign in usual dimensions , @xmath13 and @xmath14 are corresponding values for extra dimensions .
it is supposed in ref .
@xcite that the scale factor @xmath11 grows while @xmath13 diminishes , in other words , some form of dynamical compactification @xcite takes place , a size of compactified @xmath15 is small enough to play no essential role at the tev scale .
the authors of ref .
@xcite develop the approach of ref .
@xcite and suppose that the spacetime ( [ metricind ] ) is filled with a uniform density matter with anisotropic pressure and the following energy - momentum tensor : @xmath16 here @xmath17 is the energy density and @xmath18 ( @xmath19 ) is the pressure in normal ( extra ) dimensions .
so in normal dimensions pressure is different from that in additional dimensions , while being isotropic within each subspace . in ref .
@xcite matter in the form of a single fluid is supposed to behave like pressureless dust ( @xmath20 ) in usual dimensions , while in extra dimensions it has appreciable pressure @xmath19 depending on density @xmath17 by a power law @xmath21 with a negative constant @xmath22 .
the latter equation of state resembles a generalized chaplygin gas @xcite . in this model matter ( [ tmn ] ) with anisotropic pressure plays a role of dark energy and source of accelerated expansion .
so the following einstein equation without usual @xmath0 term is considered : @xmath23 to describe the late time acceleration of the universe many authors @xcite used the similar approach , in particular , extra dimensions , a metric of the type ( [ metricind ] ) and the energy - momentum tensor ( [ tmn ] ) . however , the cited authors used different equations of state . in particular , in refs .
@xcite these equations were linear @xmath24 under these conditions a set of cosmological solutions with power law dependence of @xmath25 , @xmath15 , @xmath17 on @xmath26 was obtained in refs .
but for these solutions an acceleration for @xmath25 and a dynamical compactification or stabilization for @xmath15 are not possible simultaneously .
the similar problem appears in ref .
@xcite , where the authors use the sum of two perfect fluids with densities @xmath17 and @xmath27 and the equations of state @xmath28 , @xmath29 . in this case for solutions with @xmath30 an acceleration ( @xmath31 ) suppresses any compactification or diminishing for @xmath13 .
the problem of dynamical compactification for the extra dimensions was solved in the paper by mohammedi @xcite under assumptions ( [ metricind ] ) with @xmath32 , ( [ tmn ] ) and the following ansatz : @xmath33 mohammedi constructed solutions with accelerated expansion without a predetermined equation of state . in his approach evolution of values @xmath17 , @xmath18 , @xmath19 was calculated from the right hand sides if eqs .
( [ eeq ] ) with a @xmath0 term .
relations between these values correspond to equations of state , they appear at the last stage of this scheme .
application of the mohammedi s solutions @xcite to describing the observational data will be discussed below .
middleton and stanley in ref .
@xcite in the framework of the linear equations of state ( [ eosl ] ) deduced the relation @xmath34 generalizing eq .
( [ mohans ] ) . here
they obtained a set of cosmological solutions including a hypergeometric function of powers of @xmath25 .
however , for these solutions an accelerated expansion of @xmath25 takes place only when the eos parameters @xmath36 , @xmath37 in eq .
( [ eosl ] ) are both negative , and also an accelerated expansion of @xmath25 in the late universe is incompatible with dynamical compactification of @xmath15 @xcite .
this conclusion corresponds to the findings in refs .
@xcite .
it is worth noting that the cosmological acceleration with the dynamical compactification of extra dimensions may be achieved in scalar - tensor theories , in particular , in 5-dimensional brans - dicke models @xcite .
but these models along with the extra metric component @xmath38 require the additional degree of freedom in the form of the scalar brans - dicke field @xmath39 .
this paper is organized as follows . in sec .
[ cosm ] we show , that the model @xcite not only for @xmath40 but also in the case @xmath6 can describe the current acceleration of the universe with dynamical compactification of @xmath15 . in sec .
[ obs ] we apply this model for all @xmath41 to describing observational data for type ia supernovae and determine optimal model parameters . in sec .
[ modif ] we modify the model @xcite to solve the above mentioned problem of `` the end of the world '' .
for the considered metric ( [ metricind ] ) in the case @xmath32 the einstein tensor components @xmath42 ( @xmath43 , @xmath44 ) are @xcite : @xmath45 - 3\frac{\ddot{a}a+\dot{a}^2+k_1 } { a^2}. \nonumber\end{aligned}\ ] ] if we substitute these expressions into eq .
( [ eeq ] ) and add the continuity condition @xmath46 we obtain the system of cosmological equations .
this system has the form @xmath47 in the case with @xmath6 extra spatial dimension , that did not considered in ref .
@xcite . here
pressure @xmath18 in `` usual '' dimension equals zero , as mentioned above .
( [ contin ] ) is the continuity condition for @xmath6 and @xmath20 . using the hubble constant @xmath48 c@xmath49 @xcite and the critical density @xmath50 at the present time , we make the following substitutions @xmath51 and introduce dimensionless time @xmath52 , density @xmath53 , pressure @xmath54 and logarithms @xmath55 , @xmath56 of the scale factors ( here @xmath57 , @xmath58 are present time values of @xmath25 and @xmath15 ) .
we denote derivative with respect to @xmath52 as primes and rewrite the system ( [ e00 ] ) ( [ contin ] ) as follows : @xmath59 here @xmath60 if we express @xmath61 from eq .
( [ a1 ] ) and substitute it into three equations ( [ a2 ] ) ( [ rho ] ) , one should note that eq .
( [ a2 ] ) may be reduced to eq .
( [ a3 ] ) .
so in the planar case @xmath62 we have the system of two independent equations @xmath63
if we fix an equation of state for pressure @xmath54 , for example , the above mentioned power law ( [ pbrho1 ] ) @xmath64 we may consider the equations ( [ sysa ] ) , ( [ sysrho ] ) as a closed system of first order differential equations with respect to 2 unknown functions @xmath65 and @xmath66 . the dependence ( [ pbro ] ) is used in ref .
@xcite , where parameters @xmath67 and @xmath68 are chosen in accordance with observations .
the cauchy problem for the system ( [ sysa ] ) , ( [ sysrho ] ) requires two initial conditions .
we refer them to the present epoch ( here and below it corresponds to the value @xmath69 ) in the following form : @xmath70 the first condition results from definition of the hubble constant @xmath71 in the second condition ( [ initial ] ) we suppose that the energy density @xmath72 at the present time has the fraction @xmath73 in the critical density ( [ rocr ] ) . in ref .
@xcite this fraction equals matter density fraction in the @xmath0cdm model @xcite : @xmath74 note that in ref .
@xcite the second condition ( [ initial ] ) was used in the form @xmath75 , but the value @xmath73 ( [ omega0 ] ) was taken as the factor in the r.h.s .
( [ eeq ] ) . from our point of view , that approach introduces useless vagueness in physical sense of the value @xmath17 . in our approach @xmath17 in conditions ( [ initial ] ) is density of all gravitating matter ( visible and dark ) with described above anisotropic pressure .
remind that we have no dark energy or @xmath0 term in eq .
( [ eeq ] ) in the model @xcite .
anisotropic pressure in additional dimensions plays here the role of dark energy as a source of acceleration .
the contribution of this source is the term @xmath76 in the equality @xmath77 that results from equation ( [ a1 ] ) , if we fix it at the present time @xmath69 . to obtain cosmological solutions for @xmath6 ,
@xmath78 in this model we are to solve numerically the cauchy problem for the system ( [ sysa ] ) , ( [ sysrho ] ) with initial conditions ( [ initial ] ) moving into the past for @xmath79 and into the future for @xmath80 .
then we integrate functions @xmath65 and @xmath81 ( [ b ] ) keeping in mind eqs .
( [ tau ] ) and calculate dependence of the scale factors @xmath82 , @xmath83 and density @xmath53 on dimensionless time @xmath52 .
the results of calculation for scale factors @xmath84 , @xmath85 , density @xmath66 and the acceleration parameter @xmath86 ( @xmath87 is the deceleration parameter ) are presented in fig .
here @xmath78 , @xmath88 , @xmath89 and 3 scenarios for @xmath90 ( dash - dotted line ) , @xmath91 ( solid lines ) and @xmath92 ( dashed lines ) are shown .
this evolution begins from infinite value of density @xmath53 at some initial moment @xmath93 .
we can see here two different variants for this beginning . for solutions with @xmath90 and @xmath91 ( we denote them as `` regular '' solutions ) the scale factor @xmath25 expands from @xmath94 like @xmath95 at the initial stage whereas the scale factor @xmath15 diminishes from initial infinite value up to values @xmath96 during some percent of total lifetime of this universe .
this behavior of @xmath85 looks like some variant of dynamical compactification , because the parameter @xmath58 is arbitrary one in this model , we may put @xmath58 to be sufficiently small .
another type of evolution ( `` singular '' solutions ) is represented with dashed lines in fig .
[ if1 ] for @xmath92 . for singular solutions
infinite value of density @xmath53 at @xmath97 corresponds to nonzero value of the scale factor @xmath25 and @xmath98 .
obviously , these solutions are nonphysical and should be excluded . all regular and singular solutions in fig .
[ if1 ] describe accelerated expansion ( for the factor @xmath25 ) at late stage of evolution .
beginning of this stage may be seen in the graph of the acceleration parameter @xmath99 .
acceleration rate depends on the parameters @xmath67 , @xmath68 , @xmath73 and the curvature fraction ( [ omk ] ) @xmath100 depending on the sign @xmath12 . if @xmath101 ( @xmath102 ) , one should use the system ( [ a3 ] ) ( [ b ] ) instead of eqs .
( [ sysa ] ) , ( [ sysrho ] ) . in this case
we integrate numerically the function @xmath65 simultaneously with solving the cauchy problem for the system ( [ a3 ] ) ( [ b ] ) .
we add here the natural initial condition @xmath103 to conditions ( [ initial ] ) .
for all reasonable values of four free parameters @xmath67 , @xmath68 , @xmath73 , @xmath104 the stage of accelerated expansion appears to be finite , because density @xmath53 inevitably vanishes in this model . in fig .
[ if1 ] this effect may be seen in the graphs @xmath66 with logarithmic scale in y - direction .
we denote the moment of zero density by @xmath105 : @xmath106 . for @xmath107 density @xmath53
becomes negative and nonphysical , all energy conditions ( in particular , the weak energy condition ) are violated .
this finite - time future singularity may be classified as the type iv singularity in accordance with the scheme from refs .
@xcite . for this singularity @xmath108 is nonzero , @xmath109 equals zero , the effective density and pressure @xmath110 remain nonzero , but higher derivatives of @xmath111 diverge at @xmath112 . note that the main features of the considered cosmological solutions , in particular , the future singularity , finite lifetime @xmath113 and negative density for @xmath107 take place not only for @xmath6 , but also for higher dimensions @xmath40 . in the case of @xmath40 additional dimensions after substituting the components @xmath42 into einstein equation ( [ eeq ] ) and substitutions ( [ tau ] ) in these equations and eq .
( [ contin ] ) we have in the flat case @xmath114 the following system @xcite , generalizing eqs .
( [ b ] ) ( [ sysrho ] ) : @xmath115/d}-3a'}{d-1}.\nonumber \end{aligned}\ ] ] solutions of the system ( [ sysrhod ] ) for @xmath40 were obtained in ref .
@xcite , but some features of them were not considered in that paper .
for example , singular solutions with nonzero value @xmath116 ( where @xmath53 is infinite at the initial moment @xmath93 ) also take place for @xmath40 , if the value @xmath67 is less than the critical value @xmath117 . in fig .
[ if2 ] boundaries @xmath118 separating domains of regular and singular solutions on the @xmath119 plane are presented for different @xmath7 and @xmath73 .
singular solutions are described by the inequality @xmath120 and lie below corresponding lines in fig .
[ if2 ] . between domains of regular ( above ) and singular solutions ( below a curve ) for indicated values @xmath73 and @xmath7 ] another important property of these cosmological solutions is their finite - time future singularity , in other words , inevitability of `` the end of the world '' because of vanishing density at @xmath122 for all @xmath7 ( see fig .
[ if4 ] below ) .
the authors of ref .
@xcite did not pay attention to this phenomenon , essential for their model .
it is connected with the chosen equation of state ( [ pbro ] ) for pressure @xmath54 in extra dimensions .
this drawback will be eliminated with modifying the model @xcite in sect .
[ modif ] after application this model to describing observational data for type ia supernovae in the next section .
to apply the model to describing the observational data it is convenient , following the authors of @xcite , to use internet table @xcite for type ia supernovae in distant galaxies . at the present moment
this updated table contains redshifts @xmath123 , distance moduli @xmath124 and errors @xmath125 of @xmath124 for @xmath126 supernovae .
redshift @xmath127 is associated with the value of @xmath25 at the time @xmath26 of a supernova light emission .
the distance modulus @xmath128 is the logarithmic function @xmath129 of the luminosity distance @xcite : @xmath130 to describe the data @xcite of type ia supernovae , for given values @xmath7 ,
@xmath67 , @xmath68 , @xmath73 of this model we consider evolution of the scale factor @xmath84 and dependence of the numerical integral ( [ mudl ] ) @xmath131 and @xmath128 on @xmath52 . for each value of redshift @xmath132 in the table
@xcite we calculate the corresponding @xmath133 with using eq .
( [ z ] ) and linear approximation and the theoretical value @xmath134 for @xmath135 from eq .
( [ mudl ] ) .
the measure of differences between these theoretical values @xmath136 and the measured values @xmath124 is @xcite : @xmath137 ^ 2}{\sigma_i^2}. \label{chi}\ ] ] the authors of ref .
@xcite calculated optimal parameters @xmath67 and @xmath68 , minimizing the function ( [ chi ] ) for the flat model ( @xmath78 ) with fixed @xmath88 ( [ omega0 ] ) and @xmath40 . in this approach for each @xmath40 they minimized the function @xmath138 of two variables .
we generalize their approach to the case @xmath6 additional dimension . at the first step we fix @xmath78 , @xmath88 in according with ref .
@xcite and obtain the picture of level lines for the function @xmath138 , presented in fig .
[ if3 ] for @xmath6 and @xmath139 . for @xmath78 , @xmath88 .
the dashed line is the boundary of singular solutions ] here the dashed line is taken from fig .
[ if2 ] and separates regular and singular solutions .
we see that for @xmath6 and @xmath139 the minimum of @xmath140 lies above this line , that is in the domain of regular solutions .
the same picture also takes place for @xmath141 . for each @xmath41 we calculated minimums for the function of two variables @xmath138 and coordinates @xmath67 , @xmath68 of this minimum .
they are represented in table [ tab1 ] .
.minimum of @xmath140 and optimal values @xmath67 , @xmath68 for fixed @xmath78 , @xmath88 and various @xmath7 [ cols="^,^,^,^,^,^,^",options="header " , ] [ tab3 ] we see in table [ tab3 ] that the accuracy of the model with eq .
( [ pbm ] ) increases ( @xmath140 diminishes ) for large @xmath7 , unlike in the case with eq .
( [ pbro ] ) in table [ tab2 ] .
we should note that the values @xmath140 in table [ tab3 ] are not absolutely minimal , because we fixed the parameter @xmath142 .
it is interesting , that for all @xmath7 we can achieve smaller values min@xmath140 , if we take smaller values of @xmath143 .
but if @xmath144 , the factor @xmath145 in the exponent ( [ rhoex ] ) tends to infinity , the density @xmath53 decreases too rapidly and the picture of vanishing @xmath53 looks like in the finite case in fig .
so we put the restriction @xmath146 to exclude this almost instantaneous transition to the state with @xmath147 . under this constraint
we have the optimal value @xmath142 and also @xmath148 , @xmath149 for all @xmath7 .
[ if5 ] demonstrates cosmological solutions for the model with eq .
( [ pbm ] ) with the optimal values of parameters from table [ tab3 ] . for both models
( [ pbm ] ) and ( [ pbro ] ) in figs . [ if4 ] and [ if5 ] the acceleration epoch is finite and its duration depends on @xmath7 in the same manner . but after this epoch for the model with eq .
( [ pbm ] ) we see here infinite decelerated expansion . ,
@xmath85 , functions @xmath99 and @xmath150 for the model with eq .
( [ pbm ] ) with the parameters from table [ tab3 ] ] graphs of @xmath84 , @xmath99 and @xmath150 here are more close to the dashed lines for the @xmath0cdm model during the acceleration epoch than the similar curves in fig .
but after the mentioned epoch predictions of the @xmath0cdm model and the model ( [ pbm ] ) sharply diverge . in fig .
[ if6 ] we present how the model with eq .
( [ pbm ] ) for @xmath6 , 2 , 6 and the @xmath0cdm model describe the supernovae data from the site @xcite ( dots ) in the @xmath151-@xmath131 plane .
all these models were considered below in fig .
[ if5 ] , we use the same optimal parameters from table [ tab3 ] and the same notations for the curves , in particular , the dashed line for the @xmath0cdm model . in gpc depending on redshift @xmath151 for the models @xmath0cdm and ( [ pbm ] ) with parameters from table [ tab3 ] .
dots are the data @xcite ] the presented curves are very close in the region @xmath152 , for larger @xmath151 the @xmath0cdm line slightly diverges from others .
these lines are result of optimal fitting to the observational data @xcite ( 580 dots in fig . [ if6 ] ) .
the values @xmath140 in table [ tab3 ] show rather good results for the model @xcite with eq .
( [ pbm ] ) for pressure , but these values are not the best fit , because we fixed @xmath143 to avoid the mentioned above sharp transition to @xmath147 .
the gravitational model of pahwa , choudhury and seshadri @xcite with additional spatial dimensions and anisotropic pressure provides accelerated expansion of the universe corresponding to observational data for type ia supernovae @xcite simultaneously with dynamical compactification of @xmath7 extra dimensions .
it is important that such a behavior of solutions results from rather simple equations of state ( [ pbro ] ) .
this approach is more natural in comparison with the scheme of mohammedi @xcite , where complicated equations of state are deduced from the constructed solutions , in particular , from the solution ( [ moh ] ) described below .
the authors of ref .
@xcite did not consider the case @xmath6 , but we found that for the chosen in ref .
@xcite power law equation of state ( [ pbro ] ) it is the case @xmath6 that yields the best fit ( see tables [ tab1 ] , [ tab2 ] ) .
unfortunately , the model @xcite with the power law dependence ( [ pbro ] ) inevitably predicts the finite - time future singularity of type iv in classification from refs . @xcite .
it is connected with vanishing density @xmath27 at finite time @xmath122 ( and negative density for @xmath107 ) .
evolution of this universe is broken at @xmath105 , the finite lifetime is shorter for small @xmath7 ( see fig .
[ if4 ] ) .
we demonstrated in sect .
[ modif ] , that this drawback has technical nature .
it is connected with too weak dependence @xmath153 for small @xmath27 in the power law equation of state ( [ pbro ] ) .
if we modify this law and choose a linear dependence ( [ pb0 ] ) for small @xmath27 , we obtain an infinite cosmological evolution with positive density ( but vanishing at @xmath154 ) .
we suggest the linear - fractional variant ( [ pbm ] ) of dependence @xmath155 to solve the following two problems : ( a ) to avoid `` the end of the world '' of the type ( [ rhof ] ) and ( b ) to describe 580 type ia supernovae data points from the site @xcite . the dependence ( [ pbm ] ) is a bit more complicated than eq .
( [ pbro ] ) , but it successfully conserves positive density @xmath27 during infinite lifetime and fits the data @xcite better than the @xmath0cdm model and the model from ref .
@xcite with eq .
( [ pbro ] ) .
although the simplicity of the equations of state ( [ pbrho1 ] ) is the important advantage of the model @xcite , we are to step back from this simplicity .
but in our opinion , the dependence ( [ pbm ] ) is the minimal retreat that solves this problem .
our calculations should be compared with predictions of other multidimensional models @xcite .
we shall consider the models , describing the late time acceleration of @xmath11 together with a contraction of @xmath13 , in particular , the mohammedi model in ref .
@xcite with the ansatz ( [ mohans ] ) @xmath156 .
it ensures a dynamical compactification , if @xmath157 is positive and @xmath11 expands . for @xmath157 satisfying the equality @xmath158 a set of solutions with accelerated expansion
was obtained in ref .
@xcite in the form @xmath159 here @xmath160 , the natural condition @xmath161 must be saticfied .
we mentioned above , that the equation of state in the mohammedi s approach may be determined at the last stage after substitution of the expressions ( [ moh ] ) and ( [ mohans ] ) into eqs .
( [ eeq ] ) with a @xmath0 term .
in particular , the relation between @xmath18 and @xmath17 for solutions ( [ moh ] ) results from the first two equations ( [ eeq ] ) ( in our notations ) @xmath162/a^2\rule{0mm}{1.3em}\\ -dn(n+1)\,\mu^2+\lambda=8\pi gp_a,\rule{0mm}{1.3em } \end{array}\ ] ] if we exclude @xmath163 .
this equation of state is mush more complicated than its analog @xmath20 for the model @xcite , in addition it has the negative limit of @xmath18 at @xmath164 for the case @xmath165 .
if we accept these complicated equations of state for the model @xcite , we can obtain the optimal solution ( [ moh ] ) , minimizing the sum @xmath140 ( [ chi ] ) for the same supernovae data @xcite .
for this purpose we use 2 fitting parameters of these solutions : @xmath166 and @xmath167 .
the calculations result in the optimal values @xmath168 , @xmath169 and the corresponding minimum @xmath170 .
this minimum is close to the results of the considered model @xcite in tables [ tab2 ] and [ tab3 ] .
so we may conclude that the solution ( [ moh ] ) describes the supernovae data @xcite rather successfully .
it is interesting that solutions close to eq .
( [ moh ] ) appeared in refs .
@xcite in the brane model . in refs .
@xcite darabi obtained exponential solutions @xmath171 for the model with varying @xmath172 .
such a solution with one fitting parameter is less adaptable in comparison with eq .
( [ moh ] ) , the optimal sum ( [ chi ] ) in this case @xmath173 .
note that in this paper for the model with eq .
( [ pbm ] ) we practically used only two fitting parameters @xmath67 and @xmath174 .
the value @xmath143 was fixed because for very small @xmath143 we have better fit , but the sharp downfall of @xmath150 to @xmath175 looks like the mentioned `` end of the world '' .
the parameters @xmath73 and @xmath104 influence on minimum of @xmath140 rather weakly for the model with eq .
( [ pbm ] ) . if we fix , for example , @xmath88 and @xmath176 , minimums for @xmath140 will differ from results in table [ tab3 ] less then 0.01 for all @xmath7 .
cosmological solutions in the model with eq .
( [ pbm ] ) are divided into regular and singular ones similarly to solutions with eq .
( [ pbro ] ) shown in fig .
[ if1 ] . however , fig .
[ if5 ] demonstrates that for the optimal values of parameters from table [ tab3 ] solutions with eq .
( [ pbm ] ) are regular .
it is interesting that the model @xcite with both considered variants of dependence @xmath54 on @xmath53 ( [ pbro ] ) and ( [ pbm ] ) predicts finiteness of the acceleration epoch .
its duration depends on @xmath7 in the same manner ( compare figs .
[ if4 ] and [ if5 ] ) and then acceleration sharply turns to deceleration . in the case
( [ pbro ] ) this evolution is broken at @xmath122 with @xmath177 , but for the model with eq .
( [ pbm ] ) the decelerated expansion is infinite and density @xmath150 tends to zero in the exponential form ( [ rhoex ] ) . | we consider the gravitational model with additional spatial dimensions and anisotropic pressure which is nonzero only in these dimensions .
cosmological solutions in this model include accelerated expansion of the universe at late age of its evolution and dynamical compactification of extra dimensions .
this model describes observational data for type ia supernovae on the level or better than the @xmath0cdm model .
we analyze two equations of state resulting in different predictions for further evolution , but in both variants the acceleration epoch is finite . |
our present understanding of the universe is based upon the successful hot big bang theory , which explains its evolution from the first fraction of a second to our present age , around 13 billion years later .
this theory rests upon four strong pillars , a theoretical framework based on general relativity , as put forward by albert einstein and alexander a. friedmann in the 1920s , and three strong observational facts .
first , the expansion of the universe , discovered by edwin p. hubble in the 1930s , as a recession of galaxies at a speed proportional to their distance to us .
second , the relative abundance of light elements , explained by george gamow in the 1940s , mainly that of helium , deuterium and lithium [ see fig . 1 ] , which were cooked from the nuclear reactions that took place at around a second to a few minutes after the big bang , when the universe was a hundred times hotter than the core of the sun .
third , the cosmic microwave background ( cmb ) , the afterglow of the big bang , discovered in 1965 by arno a. penzias and robert w. wilson as a very isotropic blackbody radiation at a temperature of about 3 degrees kelvin ( degrees centigrade above absolute zero ) , emitted when the universe was cold enough to form neutral atoms , and photons decoupled from matter , approximately 300000 years after the big bang .
today , these observations are confirmed to within a few percent accuracy , and have helped establish the hot big bang as the preferred model of the universe .
the big bang theory could not explain , however , the origin of matter and structure in the universe ; that is , the origin of the matter - antimatter asymmetry , without which the universe today would be filled by a uniform radiation continuosly expanding and cooling , with no traces of matter , and thus without the possibility to form gravitationally bound systems like galaxies , stars and planets that could sustain life .
moreover , the standard big bang theory assumes but can not explain the origin of the extraordinary smoothness and flatness of the universe on the very large scales , seen by the microwave background probes and the largest galaxy catalogs [ see fig . 2 ] .
it can not explain the origin of the primordial density perturbations that gave rise to cosmic structures like galaxies , clusters and superclusters , via gravitational collapse ; neither the quantity and nature of the dark matter that we believe holds the universe together ; nor the origin of the big bang itself . in the 1980s a new paradigm , deeply rooted in fundamental physics ,
was put forward by alan h. guth , andrei d. linde and others , to address these fundamental questions . according to the inflationary paradigm ,
the early universe went through a period of exponential expansion , driven by the approximately constant energy density of a scalar field called the inflaton .
in modern physics , elementary particles are represented by quantum fields , which resemble the familiar electric , magnetic and gravitational fields .
a field is simply a function of space and time whose quantum oscillations are interpreted as particles .
for instance , the photon is the particle associated with the electromagnetic field . in our case , the inflaton field has associated with it a large potential energy density , which drives the exponential expansion during inflation [ see fig . 3 ] .
we know from general relativity that the density of matter determines the expansion of the universe , but a constant energy density acts in a very peculiar way : as a repulsive force that makes any two points in space separate at exponentially large speeds .
( this does not violate the laws of causality because there is no information carried along in the expansion , it is simply the stretching of spacetime . )
this superluminal expansion is capable of explaining the large scale homogeneity of our observable universe and , in particular , why the microwave background looks so isotropic : regions separated today by more than a degree in the sky were , in fact , in causal contact before inflation , but were stretched to cosmological distances by the expansion [ see fig .
any inhomogeneities present before the tremendous expansion would be washed out . moreover , in the usual big bang scenario a flat universe , one in which the gravitational attraction of matter is exactly balanced by the cosmic expansion , is unstable under perturbations : a small deviation from flatness is amplified and soon produces either an empty universe or a collapsed one . for the universe to be nearly flat today , it must have been extremely flat at nucleosynthesis for example , deviations not exceeding more than one part in @xmath0 .
this extreme fine tuning of initial conditions was also solved by the inflationary paradigm [ see fig . 5 ]
. thus inflation is an extremely elegant hypothesis that explains how a region much , much greater that our own observable universe could have become smooth and flat without recourse to _ ad hoc _ initial conditions .
if cosmological inflation made the universe so extremely flat and homogeneous , where did the galaxies and clusters of galaxies come from ?
one of the most astonishing predictions of inflation , one that was not even expected , is that quantum fluctuations of the inflaton field are stretched by the exponential expansion and generate large scale perturbations in the metric .
inflaton fluctuations are small wave - packets of energy that , according to general relativity , modify the spacetime fabric , creating a whole spectrum of curvature perturbations .
the use of the word spectrum here is closely related to the case of light waves propagating in a medium : a spectrum characterises the amplitude of each given wavelength . in the case of inflation ,
the inflaton fluctuations induce waves in the spacetime metric that can be decomposed into different wavelengths , all with approximately the same amplitude , that is , corresponding to a scale - invariant spectrum .
these patterns of perturbations in the metric are like fingerprints that characterise unequivocally a period of inflation .
when matter fell in the troughs of these waves , it created density perturbations that collapsed gravitationally to form galaxies , clusters and superclusters of galaxies , with a spectrum that is also scale - invariant .
such a type of spectrum was proposed in the early 1970s ( before inflation ) by edward r. harrison , and independently by the russian cosmologist yakov b. zeldovich , to explain the distribution of galaxies and clusters of galaxies on very large scales in our observable universe . various telescopes , like the hubble space telescope
, the twin keck telescopes in hawaii and the european southern observatory telescopes in chile , are exploring the most distant regions of the universe and discovering the first galaxies at large distances . according to the big bang theory ,
the further the galaxy is , the larger its recession velocity , and the larger the shift towards the red of the spectrum of light from that galaxy .
astronomers thus measure distances in units of red - shift @xmath1 .
the furthest galaxies observed so far are at redshifts of @xmath2 , or 12 billion light - years from the earth , whose light was emitted when the universe had only about 5 percent of its present age .
only a few galaxies are known at those redshifts , but there are at present various catalogs like the cfa and apm galaxy catalogs , and more recently the iras pscz [ see fig . 2 ] and las campanas redshift surveys , that study the spatial distribution of hundreds of thousands of galaxies up to distances of a billion light - years , or @xmath3 , that receed from us at speeds of tens of thousands of kilometres per second .
these catalogs are telling us about the evolution of clusters of galaxies in the universe , and already put constraints on the theory of structure formation based on the gravitational collapse of the small inhomogeneities produced during inflation . from these observations one can infer that most galaxies formed at redshifts of order 2 to 4 ; clusters of galaxies formed at redshifts of order one , and superclusters are forming now .
that is , cosmic structure formed from the bottom up , from galaxies to clusters to superclusters , and not the other way around .
this fundamental difference is an indication of the type of matter that gave rise to structure .
we know from primordial nucleosynthesis that all the baryons in the universe can not account for the observed amount of matter , so there must be some extra matter ( dark since we do nt see it ) to account for its gravitational pull . whether it is relativistic ( hot ) or non - relativistic ( cold ) could be inferred from observations : relativistic particles tend to diffuse from one concentration of matter to another , thus transferring energy among them and preventing the growth of structure on small scales .
this is excluded by observations , so we conclude that most of the matter responsible for structure formation must be cold . how much there is is a matter of debate at the moment .
some recent analyses suggest that there is not enough cold dark matter to reach the critical density required to make the universe flat .
some other form of energy permeates the universe , if we want to make sense of the present observations . in order to resolve this issue ,
even deeper galaxy redshift catalogs are underway , looking at millions of galaxies , like the sloan digital sky survey ( sdss ) and the anglo - australian two degree field ( 2df ) galaxy redshift survey , which are at this moment taking data , up to redshifts of @xmath4 , or several billion light - years away , over a large region of the sky .
these important observations will help astronomers determine the nature of the dark matter and test the validity of the models of structure formation .
[ table1 ] .the parameters of the standard cosmological model [ cols="<,<,<,<,<",options="header " , ] everything we know about the properties of elementary particles is included in the standard model of particle physics . it describes more than a hundred observed particles and their interactions in terms of a few fundamental constituents : six quarks and six leptons , and their antiparticles [ see table 2 ] .
the standard model describes three types of interactions : the electromagnetic force , the strong and the weak nuclear forces .
these forces are transmitted by the corresponding particles : the photon , the gluon and the w and z bosons .
the theory also requires a scalar particle , the higgs particle , responsible for the masses of quarks and leptons and the breaking of the electroweak symmetry at an energy scale 100 times the mass of the proton .
the higgs is believed to lie behind most of the mysteries of the standard model , including the asymmetry between matter and antimatter .
symmetries are fundamental properties of any physical theory .
a theory is symmetric under certain symmetry operation , like reflection , if its laws apply equally well after such an operation is performed on part of the physical system .
an important example is the operation called parity reversal , denoted by p. it produces a mirror reflection of an object and rotates it 180 degrees about an axis perpendicular to the mirror .
a theory has p symmetry if the laws of physics are the same in the real and the parity - reversed world .
particles such as leptons and quarks can be classified as right- or left - handed depending on the sense of their internal rotation , or spin , around their direction of motion .
if p symmetry holds , right - handed particles behave exactly like left - handed ones
. the laws of electrodynamics and the strong interactions are the same in a parity - reflected universe .
but , as chien - shiung wu discovered in 1957 , the weak interaction acts very differently on particles with different handedness : only left - handed particles can decay by means of the weak interaction , not right - handed ones . moreover ,
as far as we know , there are no right - handed neutrinos , only left - handed .
so the weak force violates p. another basic symmetry of nature is charge conjugation , denoted by c. this operation changes the quantum numbers of every particle into those of its antiparticle .
charge symmetry is also violated by the weak interactions : antineutrinos are not left - handed , only right - handed . combining c and p one
gets the charge - parity symmetry cp , which turns all particles into their antiparticles and also reverses their handedness : left - handed neutrinos become right - handed antineutrinos [ see fig . 9 ] .
although charge and parity symmetry are individually broken by the weak interaction , one expects their combination to be conserved .
however , in 1964 , a groundbreaking experiment by james cronin , val finch and ren turlay at brookhaven national laboratory showed that cp was in fact violated to one part in 1000 .
it was hard to see why cp symmetry should be broken at all and even more difficult to understand why the breaking was so small .
soon after , in 1972 , makoto kobayashi and toshihide maskawa showed that cp could be violated within the standard model if three or more generations of quarks existed , because of cp nonconserving phases that could not be rotated away .
only two generations where known at the time , but in 1975 martin l. perl and collaborators discovered the tau lepton at the stanford linear accelerator centre ( slac ) , the first ingredient of the third generation .
only recently the last quark in the family was discovered at fermilab , the top quark .
but how does this picture fit in the evolution of the universe ? in 1967 , the russian physicist andrei sakharov pointed out the three necessary conditions for the baryon asymmetry of the universe to develop .
first , we need interactions that do not conserve baryon number b , otherwise no asymmetry could be produced in the first place .
second , c and cp symmetry must be violated , in order to differentiate between matter and antimatter , otherwise b nonconserving interactions would produce baryons and antibaryons at the same rate , thus maintaining zero net baryon number .
third , these processes should occur out of thermal equilibrium , otherwise particles and antiparticles , which have the same mass , would have equal occupation numbers and would be produced at the same rate .
the standard model is baryon symmetric at the classical level , but violates b at the quantum level , through the chiral anomaly .
electroweak interactions violate c and cp , but the magnitude of the latter is clearly insufficient to account for the observed baryon asymmetry .
this failure suggests that there must be other sources of cp violation in nature , and thus the standard model of particle physics is probably incomplete .
one of the most popular extensions of the standard model includes a new symmetry called supersymmetry , which relates bosons ( particles that mediate interactions ) with fermions ( the constituents of matter ) .
those extensions generically predict other sources of cp violation coming from new interactions at scales above 1000 times the mass of the proton .
such scales will soon be explored in the first years of the next millenium by particle colliders like the large hadron collider ( lhc ) at cern , the european centre for particle physics , and by the tevatron at fermilab .
the mechanism for baryon production in the early universe in these models relies on the strength of the electroweak phase transition , as the universe cooled and the symmetry was broken . only for strongly first order phase transitions
is the universe sufficiently far from equilibrium to produce enough baryon asymmetry .
unfortunately , the phase transition in these models is invariably too weak to account for the observed asymmetry , so some other mechanism is needed . if reheating after inflation occurred in an explosive way , via the resonant production of particles from the inflaton decay , as recent developments suggest , then the universe has actually gone through a very non - linear , non - perturbative and very far from equilibrium stage , before thermalising via particle interactions .
electroweak baryogenesis could then take place during that epoch , soon after the end of low energy inflation at the electroweak scale .
such models can be constructed but require a specially flat direction ( a very small mass for the inflaton ) during inflation , in order to satisfy the constraints from the amplitude of temperature anisotropies seen by cobe .
such flat directions are generic in supersymmetric extensions of the standard model . after inflation ,
the inflaton acquires a large mass from its interaction with the higgs field .
the crucial ingredient of departure from equilibrium , necessary for the excess production of baryons over antibaryons , is strongly present in this new scenario of baryogenesis , as the universe develops from a zero temperature and zero entropy state , at the end of inflation , to a thermal state with exponentially large numbers of particles , the origin of the standard hot big bang .
if , during this stage , fundamental or effective interactions that are b , c and cp violating were fast enough compared to the rate of expansion , the universe could have ended with the observed baryon asymmetry of one part in @xmath5 , or one baryon per @xmath6 photons today , as deduced from observations of the light element abundances [ see fig . 1 ] .
recent calculations suggest than indeed , the required asymmetry could be produced as long as some new physics , just above the electroweak symmetry breaking scale , induces a new effective cp violating interaction .
these new phenomena necessarily involve an interaction between the higgs particle , responsible for the electroweak symmetry breaking , and the inflaton field , responsible for the period of cosmological inflation .
therefore , for this scenario to work , it is expected that both the higgs and the inflaton particles be discovered at the future particle physics colliders like the lhc and the next linear collider ( nlc ) , to be built in the next millenium .
furthermore , this new physics would necessarily involve new interactions in the quark sector , for example inducing cp violations in the b meson ( a bound state composed of a bottom quark and an antidown quark ) system .
such violations are the main research objective of the b factory at slac in california and at kek , the high energy accelerator research organisation in tsukuba , japan .
those experiments will be taking data in a few months time , at the turn of the millenium , and could give us clues on the issues of the matter - antimatter asymmetry and thus on baryogenesis from reheating after inflation .
if confirmed , such a new scenario of baryogenesis would represent a leap forward in our understanding of the universe from the unifying paradigm of inflationary cosmology .
furthermore , it would bring inflation down to a scale where present or future particle physics experiments would be able to explore it quite thoroughly .
cosmological inflation thus enters the realm of testable low energy particle physics .
we have entered a new era in cosmology , were a host of high precision measurements are already posing challenges to our understanding of the universe : the density of ordinary matter and the total amount of energy in the universe ; the microwave background anisotropies on a fine scale resolution ; primordial deuterium abundance from quasar absorption lines ; the acceleration parameter of the universe from high - redshift supernovae observations ; the rate of expansion from gravitational lensing ; large scale structure measurements of the distribution of galaxies and their evolution , and many more , which already put constraints on the parameter space of cosmological models [ see fig . 10 ] .
however , these are only the forerunners of the precision era in cosmology that will dominate the new millenium , and will make cosmology a phenomenological science .
it is important to bear in mind that all physical theories are approximations of reality that can fail if pushed too far .
physical science advances by incorporating earlier theories that are experimentally supported into larger , more encompassing frameworks .
the standard big bang theory is supported by a wealth of evidence , nobody really doubts its validity anymore .
however , in the last decade it has been incorporated into the larger picture of cosmological inflation , which has become the new standard cosmological model .
all cosmological issues are now formulated in the context of the inflationary cosmology .
it is the best explanation we have at the moment for the increasing set of cosmological observations . in the next few years
we will have an even larger set of high - quality observations that will test inflation and the cold dark matter paradigm of structure formation , and determine most of the 12 or more parameters of the standard cosmological model to a few percent accuracy [ see table . 1 ] .
it may seem that with such a large number of parameters one can fit almost anything .
however , that is not the case when there is enough quantity and quality of data .
an illustrative example is the standard model of particle physics , with around 21 parameters and a host of precise measurements from particle accelerators all over the world .
this model is nowadays rigurously tested , and its parameters measured to a precision of better than a percent in some cases .
it is clear that high precision measurements will make the standard model of cosmology as robust as that of particle physics .
in fact , it has been the technological advances of particle physics detectors that are mainly responsible for the burst of new data coming from cosmological observations .
this is definitely a very healthy field , but there is still a lot to do . with the advent of better and larger precision experiments , cosmology is becoming a mature science , where speculation has given way to phenomenology .
there are still many unanswered fundamental questions in this emerging picture of cosmology .
for instance , we still do not know the nature of the inflaton field , is it some new fundamental scalar field in the electroweak symmetry breaking sector , or is it just some effective description of a more fundamental high energy interaction ?
hopefully , in the near future , experiments in particle physics might give us a clue to its nature .
inflation had its original inspiration in the higgs field , the scalar field supposed to be responsible for the masses of elementary particles ( quarks and leptons ) and the breaking of the electroweak symmetry .
such a field has not been found yet , and its discovery at the future particle colliders would help understand one of the truly fundamental problems in physics , the origin of masses . if the experiments discover something completely new and unexpected , it would automatically affect inflation at a fundamental level .
one of the most difficult challenges that the new cosmology will have to face is understanding the origin of the cosmological constant , if indeed it is confirmed by independent sets of observations . ever since einstein introduced it as a way to counteract gravitational attraction , it has haunted cosmologists and particle physicists for decades .
we still do not have a mechanism to explain its extraordinarily small value , 120 orders of magnitude below what is predicted by quantum physics . for several decades
there has been the reasonable speculation that this fundamental problem may be related to the quantisation of gravity .
general relativity is a classical theory of spacetime , and it has proved particularly difficult to construct a consistent quantum theory of gravity , since it involves fundamental issues like causality and the nature of spacetime itself . the value of the cosmological constant predicted by quantum physics is related to our lack of understanding of gravity at the microscopic level .
however , its effect is dominant at the very largest scales of clusters or superclusters of galaxies , on truly macroscopic scales .
this hints at what is known in quantum theory as an anomaly , a quantum phenomenon relating both ultraviolet ( microscopic ) and infrared ( macroscopic ) divergences .
we can speculate that perhaps general relativity is not the correct description of gravity on the very largest scales .
in fact , it has been only in the last few billion years that the observable universe has become large enough that these global effects could be noticeable . in its infancy , the universe was much smaller than it is now , and presumably general relativity gave a correct description of its evolution , as confirmed by the successes of the standard big bang theory . as it expanded , larger and larger regions were encompassed and therefore deviations from general relativity would slowly become important .
it may well be that the recent determination of a cosmological constant from observations of supernovae at high redshifts is hinting at a fundamental misunderstanding of gravity on the very large scales .
if this were indeed the case , we should expect that the new generation of precise cosmological observations will not only affect our cosmological model of the universe but also a more fundamental description of nature . | early universe ; inflation ; cosmic microwave background ; cosmological parameters ; large scale structure ; gravitational waves ; baryogenesis ; dark matter cosmology is nowadays one of the most active areas of research in fundamental science .
we are going through a true revolution in the observations that are capable of providing crucial information about the origin and evolution of the universe . in the first years of the next millenium we will have , for the first time in the history of such an ancient science as cosmology , a precise knowledge about a handful of parameters that determine our standard cosmological model .
this standard model is based on the inflationary paradigm , a period of exponential expansion in the early universe responsible for the large scale homogeneity and flatness of our observable patch of the universe .
a spectrum of density perturbations , seen in the microwave background as temperature anisotropies , could have been produced during inflation from quantum fluctuations that were stretched to cosmological size by the expansion , and later gave rise , via gravitational collapse , to the observed large scale structure of clusters and superclusters of galaxies .
furthermore , the same theory predicts that all the matter and radiation in the universe today originated at the end of inflation from an explosive production of particles that could also have been the origin of the present baryon asymmetry , before the universe reached thermal equilibrium at a very large temperature . from there on , the universe cooled down as it expanded , in the way described by the standard hot big bang model . with the observations that will soon become available in the next millenium , we will be able to test the validity of the inflationary paradigm , and determine with unprecedented accuracy the parameters of a truly standard model of cosmology .
epsf [ firstpage ] |
quantum simulators , as first proposed by feynman @xcite , which are devices built to evolve according to a postulated quantum hamiltonian and thus `` compute '' its properties , are one of the hot ideas which may provide a breakthrough in many - body physics . while one must be aware of possible difficulties ( see , e.g. , @xcite ) , impressive progress has been achieved in recent years in different systems employing cold atoms and molecules , nuclear magnetic resonance , superconducting qubits , and ions .
the latter are extremely well controlled and already it has been demonstrated that , indeed , quantum spin systems may be simulated with cold - ion setups @xcite .
quantum spin systems on a lattice constitute some of the most relevant cases for quantum simulations as there are many instances where standard numerical techniques to compute their dynamics or even their static behavior fail , especially for two- or three - dimensional systems .
chronologically the first proposition to use trapped ions to simulate lattice spin models came from porras and cirac @xcite who derived the effective spin - hamiltonian for the system , @xmath1-\mu\sum\limits_{i } s_{i}^{z } , \label{hamspin}\end{aligned}\ ] ] where @xmath2 is the chemical potential ( which in this case acts as an external magnetic field ) , @xmath3 is the interaction strength , and @xmath4 are the spin operators at site @xmath5 .
all the long - ranged interactions fall off with a @xmath6 dipolar decay , which for the ions is due to the fact that they are both induced by the same mechanism , namely lattice vibrations mediated by the coulomb force @xcite .
hamiltonian is a dipolar xxz model where the ratio of hopping to dipolar repulsion can be scaled by @xmath7 @xcite .
another route to simulate quantum magnetism of dipolar systems , using the rotational structure of the ultracold polar molecules , has been proposed in refs .
@xcite . in this case , the xy ( or tunneling ) term is restricted to nearest neighbors ( nn ) and only the zz ( or interaction ) term is dipolar .
precisely this long - range ( lr ) dipolar character makes this system interesting and challenging .
dipolar interactions introduce new physics to conventional short - range ( sr ) systems .
for example , for soft - core bosons , the system with long - range interactions and short - range tunneling , can apart from mott - insulating ( mi ) , superfluid ( sf ) , and crystal phases host a haldane - insulating phase @xcite .
this is characterized by antiferromagnetic ( afm ) order between empty sites and sites with double occupancy , with an arbitrarily long string of sites with unit occupancy in between .
another case where dipolar interactions play a crucial role is the appearance of the celebrated supersolid .
the extended bose - hubbard model ( i.e. , with nn interactions ) with soft - core on - site interactions shows stable supersolidity in one - dimensional ( 1d ) @xcite and square lattices @xcite , where in the 1d case a continuous transition to the supersolid phase occurs , in contrast to the first - order phase transition in two - dimensional ( 2d ) lattices .
long - ranged dipolar interactions on the other hand allow for the appearance of supersolidity even in the hard - core limit @xcite . in this limit ,
dipolar interactions give rise to a large number of metastable states @xcite ( for a review see @xcite ) . by tuning the direction of the dipoles , incompressible regions like devil s staircase structures
have been predicted in ref .
@xcite . while being interesting , long - range interactions make computer simulations of such systems quite difficult . on the other hand ,
since ions in optical lattices may be extremely well controlled , they form an ideal medium for a quantum simulator . in our previous paper @xcite
, we considered both the mean - field phase diagram for the system ( [ hamspin ] ) as well as a 1d chain using different quasi - exact techniques , such as density matrix renormalization group ( dmrg ) , and exact diagonalization for small systems .
here , we would like to concentrate on frustration effects on a 2d triangular lattice using a quantum monte carlo ( qmc ) approach .
for the system consisting of spin - half particles ( as , e.g. , originating from two internal ion states in the porras
cirac @xcite proposal ) the spins can be mapped onto a system of hard - core bosons , using the holstein
primakoff transformation , @xmath8 , @xmath9 , and @xmath10 , where @xmath11 @xmath12 are the creation ( annihilation ) operators of hard - core bosons and @xmath13 is the number operator at site @xmath5 .
these bosons obey the normal bosonic commutation relationships , @xmath14=1 $ ] , but are constrained to only single filling , @xmath15 .
this is the limit when the on - site repulsion term @xmath16 goes to infinity in the standard bose - hubbard model . in this representation , a spin up particle
is represented by a filled site and a spin down particle by an empty site . for convenience , we will mostly use the language of hard - core bosons in this paper . before presenting the results for the 2d triangular lattice ,
let us briefly review the behavior of the system for a 1d chain of bosons as obtained in ref .
in order to form some intuition about possible physical effects due to the lr interaction and tunneling , we now discuss briefly the ground - state phase diagram of the hamiltonian in 1d . for hard - core bosons , which is the topic of this article ,
the one - dimensional version of the hamiltonian has been thoroughly investigated in the past @xcite ( see also @xcite for the special cases @xmath17 and @xmath18 ) . for reference ,
we reproduce in fig . [ fig : phasediagram1d ] the phase diagrams for the system with dipolar interaction and nn tunneling , as well as the system with both the interaction and the tunneling terms dipolar . phase diagram of the 1d system with dipolar interactions and * ( a ) * nearest - neighbor tunneling , * ( b ) * dipolar tunneling . the phases are labeled according to density matrix renormalization results , while the actual data shown comes from the infinite time evolving block decimation method ( with interactions truncated at next - to - nearest - neighbors ) , both from ref . @xcite . along the black line
, there exists a devil s staircase of crystal states . at finite tunneling , for ( a )
, these spread into conventional insulating states , while for ( b ) , they form quasi - supersolids . the light blue lines near the center of the plots
sketch the crystal lobes at @xmath19-filling .
their cusp - like structure is typical for 1d systems . in 2d , they are expected to be rounded off , similar to mott - lobes of the bose hubbard model @xcite . for nearest - neighbor tunneling , ( a ) , the superfluid ( sf ) phases can be mapped into one another , while for nearest - neighbor dipolar tunneling , ( b ) , they are distinct on the ferromagnetic ( fm , @xmath20 ) and the antiferromagnetic side ( afm , @xmath21 ) . note also that in ( b ) frustration leads to an asymmetry between @xmath20 and @xmath21 . ] at zero tunneling , the ground states are periodic crystals where to minimize the dipolar interaction energy occupied sites are as far apart as possible for a given filling factor @xcite . for finite 1d systems and very small
tunneling such a situation persists as exemplified in @xcite . for infinite chains in 1d , every fractional filling factor @xmath22 is a stable ground state for a portion of @xmath2 parameter space .
the extent in @xmath2 decreases with @xmath23 , since at large distances the dipolar repulsion is weak and thus can not efficiently stabilize crystals with a large period .
this succession of crystal states is termed the devil s staircase .
this name derives from its surprising mathematical properties , challenging naive intuitions about continuity and measure : since all rational fillings are present , it is a continuous function ; moreover , its derivative vanishes almost everywhere ( i.e. , it is non - zero only on a set of measure zero ) and still it is not a constant , but covers a finite range . at finite tunneling , the crystals spread into lobes similar to the mott lobes of the bose
hubbard model .
if the tunneling is only over nns , these mott lobes are not sensitive to the sign of the tunneling and they form standard insulating states with diagonal long - range order ( lro ) and off - diagonal short - range order ( fig .
[ fig : phasediagram1d]a ) . for long - range dipolar tunneling ,
the extent of the lobes is asymmetric under sign change @xmath24 : frustration effects stabilize the crystal states for @xmath20 , while the ferromagnetic ( fm ) tunneling for @xmath21 destabilizes them ( fig .
[ fig : phasediagram1d]b ) .
moreover , the crystal states acquire off - diagonal correlations which follow the algebraic decay of the dipolar interactions @xcite .
this coexistence of diagonal and off - diagonal ( quasi-)lro turns the crystal states into ( quasi-)supersolids . the occurrence of such phases for hard - core bosons in 1d is truly exceptional , since the systems where it appears consist typically of soft - core bosons @xcite , or two - dimensional lattices @xcite .
furthermore , this 1d quasi - supersolid defies luttinger - liquid theory , which typically describes 1d systems very well , even in the presence of dipolar interactions @xcite . where luttinger - liquid theory applies , diagonal and off - diagonal correlations decay algebraically with exponents which are the inverse of one another .
therefore , if the diagonal correlations show lro , the corresponding exponent is effectively @xmath25 , and the exponent for the off - diagonal correlations is infinite , describing an exponential decay . in our case , this exponent remains finite in the quasi - supersolid phase and the above relationship clearly does not hold . at even stronger tunneling strengths , the crystal melts and
the system is in a sf phase .
the lr tunneling and interactions influence the correlation functions at large distances and therefore also modify the character of this phase @xcite .
these results show that in this system the dipolar interactions considerably modify the quantum - mechanical phase diagram . in higher dimensions
, we can expect the influence of long - range interactions to be even stronger , which makes extending these studies to a two - dimensional lattice highly relevant .
for example , one can expect that if quasi - supersolids appear already in 1d the long - range tunneling has a profound effect on the stability of two - dimensional supersolids , which appear in triangular lattices at the transition between crystal and superfluid phases @xcite .
also , the frustration effects already observed in the 1d system should be much more pronounced in the triangular lattice , simply due to the increased number of interactions , our method of choice , quantum monte carlo , fails due to the sign problem , invoked by frustration .
see section [ cha : results ] for details . ] .
further , such an analysis is especially relevant at finite temperature .
in fact , a recent work scanned the phase diagram of hamiltonian along the line @xmath26 in a square lattice @xcite .
there , the authors found that above the superfluid on the fm side ( i.e. , @xmath20 ) , the continuous u(1 ) symmetry of the off - diagonal correlations remains broken even at finite temperatures .
thus , the long - range nature of the tunneling leads to a phase which defies the mermin
wagner theorem @xcite . for these reasons
, we can expect intriguing effects of the long - range tunneling for the two dimensional triangular lattice , to which we turn now .
to analyze the system of dipolar bosons on the two dimensional triangular lattice , we employ quantum monte carlo ( qmc ) simulations at a finite but low temperature .
specifically , we study rhombic lattices with periodic boundary conditions and @xmath27 sites , with @xmath28 .
we will in particular thoroughly investigate the wigner crystal at @xmath19-filling .
the triangular lattice is frustrated even with only nn interactions .
when any longer ranged interactions are added , this only increases the frustration , which also makes qmc calculations more difficult . also , in frustrated systems @xcite and systems with long - range interactions @xcite typically many metastable states appear , making finding the ground state a somewhat difficult task . to avoid complications with the sign problem , caused by negative probabilities in qmc codes , we will take only negative values of @xmath7 under consideration .
note that this sign problem appears only for the long - ranged xy interactions , while the zz ( ising - like ) interactions are despite frustration
sign - problem free . in this study
, we compare the hamiltonian with both long - range interactions and hoppings ( lr lr ) , with long - ranged dipolar interactions but nn hopping ( lr sr ; relevant for polar molecules @xcite ) , as well as with hopping and interactions truncated at nns ( sr sr ; this is the nn xxz model , relevant to magnetic materials with planar anisotropy in their couplings ) , this amounts to including interactions up to distances of the fifth - nearest neighbor . ] .
each of these systems will display different crystal , superfluid , and supersolid regions .
comparing these cases will give valuable insight into the influence of the long - range terms .
all the calculations were performed using the worm algorithm of the open source alps ( algorithms and libraries for physics simulations ) project @xcite .
this algorithm , first created by n. prokofev , works by sampling world lines in the path integral representation of the partition function in the grand canonical ensemble @xcite . the first calculation , which creates the motivation for the rest of the paper , is to look at the case of vanishing tunneling and temperature for each system ( lr lr , lr sr , and sr
sr ) . here ,
similar to the 1d devil s staircase , at vanishing temperature a series of insulating crystal states is expected to cover the entire range of @xmath29 .
since we are interested in finite temperature results , we set @xmath30 which should still be low enough to reflect the characteristics of the ground - state phase diagram and look for plateaus in the density .
we distinguish short- and long - ranged zz interactions . from fig .
[ devil ] , left panel , we can see that the only plateau ( besides the completely filled system ) that appears is at @xmath31 ( corresponding either to @xmath19 boson filling or in spin terms , a lattice with @xmath19 of the spins oriented up and @xmath32 oriented down ) for both short- and long - ranged interactions . scaling the system size from @xmath33 to @xmath34
causes no change for the short - ranged interactions , and minimal change for the long - ranged ones .
the key difference is in the size and position of the short - ranged and the long - ranged plateaus . for short - ranged interactions
this plateau is larger and centered around @xmath35 , while the long - ranged interactions have a smaller plateau centered around @xmath36 .
the finite width of these plateaus suggests that a @xmath19-filling wigner crystal persists also for some finite @xmath7 .
the right panels of fig . [ devil ] show how the plateau shrinks with increasing temperature , as well for sr interactions ( top right panel ) as for lr interactions ( bottom right panel ) . in the latter case ,
in fact , by @xmath37 the plateau has completely disappeared .
we can also notice that at @xmath38 there are signs of some of the other plateaus , most noticably the @xmath39-filling plateau .
the rest of the paper will focus on the @xmath19-filling crystal lobes and their properties . and @xmath30 , where the density shows a single plateau for @xmath19-filling .
for sr interactions ( solid blue line ) , different curves for @xmath40coincide , and for lr interactions ( solid red : @xmath33 , dashed red : @xmath41 and @xmath42 ) the size dependence is small .
the panels on the right are at fixed @xmath34 and @xmath17 for different temperatures @xmath43 ( from dotted to dashed to solid ) . both for sr ( top right ) and lr interactions ( bottom right ) , the @xmath19-filling plateau shrinks as t increases . , scaledwidth=100.0% ] we now introduce a finite tunneling by choosing @xmath20 in our hamiltonian , and study the properties around the @xmath19-filling crystal .
we calculate the density and the superfluid fraction .
the superfluidity is measured using the winding numbers calculated from the movement of the worms in the qmc code . in order to get this value the system must have periodic boundary conditions so that the world lines can properly `` wind '' around the system .
the superfluid fraction is @xmath44 where @xmath45 is the winding number fluctuation of the world lines and @xmath46 is the inverse temperature .
figure [ lobes ] shows the results for the boson density of a @xmath33 triangular lattice at @xmath30 .
for all types of interactions we see that as @xmath7 increases in absolute value the @xmath47 plateau shrinks , because larger @xmath48 increases the ratio of hopping to dipolar interactions .
this introduces more kinetic energy and the crystal melts into a superfluid .
it can be seen for the sr sr system that the boson density lobe extends to @xmath49 , while for the lr sr interactions it ends at @xmath50 , and finally for lr lr interactions the lobe is smaller still , only going out to @xmath51 .
the behavior is explained by the fact that the increased amount of interactions cause a quicker melting of the lobe .
lobes , evidenced by particle density ( top row ) and the superfluid fraction ( bottom row ) that arise under varying the ratio , @xmath7 , and the chemical potential , @xmath29 ( data for @xmath33 ) . the left column corresponds to the sr
sr system , the middle one shows the lr sr system and the right column depicts the lobes for the lr lr system .
long - ranged dipolar interactions decrease the lobes in @xmath29 due to the appearance of devil s staircase like features , and long - range tunneling decreases the extent of the lobe in @xmath7.,scaledwidth=100.0% ] a second major observation is the shift in the position of the lobes .
while the short - ranged lobe exists approximately for @xmath52 , the long - ranged lobes lie generally on the interval @xmath53 . for a system like this at @xmath54 ,
one expects the @xmath47 lobe to exist on the range @xmath55 with a mirrored image of the @xmath56 lobe at @xmath57 .
the existence of the two identical lobes is explained by particle - hole symmetry @xcite .
the lobes are separated with a kind of mixed solid in between ( with coexistence of @xmath32- and @xmath19-filling regions ) the existence of this region may be caused by several phenomena .
it could either be an effect of the finite temperature as in ref .
@xcite , or due to the existence of many metastable states caused by a devil s staircase like behavior , similar to what was observed in ref .
@xcite .
again referring to the @xmath54 phase diagram , we expect that there is a region of supersolidity that extends in between the lobes and goes all the way to their base at @xmath58 @xcite . in our system
this region should exist near the tip of the lobe but not extend all the way to the base due to the finite temperature and the resulting mixed solid . looking at fig .
[ lobes ] , it is obvious indeed that , if a supersolid region exists , it can only be near the tip of the lobe because the superfluidity is zero a significant way up the lobe . judging by the increased separation of the long - ranged lobes , we can assume that the supersolid region for these systems should increase in size to fill the region in between . to search for the supersolid phase
, we now compare the superfluidity with the static structure factor .
the structure factor is defined as the fourier transform of the density - density correlations , @xmath59 here , we focus on the wave vector @xmath60 , which corresponds to the @xmath61 order parameter that is associated with @xmath32- and @xmath19-filling crystals on the triangular lattice . for the case of the @xmath19-filling lobe
that we are interested in , it will show plateaus over the same range of @xmath2 as the density , but additionally gives insight into the arrangement of the bosons on the lattice .
this makes it a useful quantity in searching for supersolid regions .
in fact , a supersolid exists when both the structure factor and the superfluid fraction have non - zero values .
the physical mechanism behind the supersolid phenomenon is based upon the appearance of extra holes ( particles ) .
the underlying crystal structure has @xmath61 order on a triangular sublattice of the physical lattice .
the extra holes ( particles ) are free to move around on the rest of the lattice as superfluid objects . in this way
, the system retains a crystal structure , while it acquires at the same time the long - range coherence of a superfluid . due to the hole ( particle ) doping , it forms in sections away from commensurate filling , in this case in between the @xmath32- and @xmath19-filling lobes . for sr
sr ( left top ) , lr sr ( left bottom ) and lr lr ( right ) for a @xmath62 , and @xmath42 lattice ( lines become thicker and darker with increasing system size ) .
for all cases , the structure factor ( solid blue ) is finite at small @xmath7 and the superfluid fraction ( dashed black ) at large @xmath7 . at
the system sizes studied , there appears a supersolid region at intermediate @xmath7 where both structure factor and superfluid fraction are finite . in the lr lr system there is a reversal of finite size effects . in this case
the superfluid fraction for larger systems becomes higher instead of lower.,scaledwidth=100.0% ] taking `` slices '' out of the crystal lobes we now check where there is a supersolid region and where the system transitions directly from crystal to superfluid .
also we study the nature of these transitions to see if they are of first or second order .
the most logical region to look for supersolids is directly in between the @xmath32- and @xmath19-filling lobes , at @xmath63 .
figure [ horizss ] shows the behavior of the sr
sr , the lr sr , and the lr lr system at this cut for several system sizes . structure factor and superfluidity reveal , for all the systems , three different regions . in each case , the system starts at @xmath17 in a solid phase where the superfluid fraction is zero but the structure factor is finite .
it transitions smoothly into a supersolid region where both superfluidity and structure factor are non - zero .
finally , the structure factor smoothly drops away and leaves just a non - zero superfluid fraction , making the final phase a superfluid . in each system
, the size of the supersolid region is different . in the sr
sr case , the supersolid region begins to appear at @xmath64 for a @xmath33 lattice . as the size grows to @xmath34 ,
the region has shifted to @xmath65 with the superfluid curves becoming sharper .
the increased system size also reduces the value of the structure factor a little . from @xcite , we know that at even larger sizes ( but @xmath54 ) the supersolid will continue to exist in this type of system .
for the lr sr system the supersolid region appears at a similar point and also shifts with system - size increase .
the structure factor on the other hand has a significant decrease for larger system sizes .
it is difficult to tell if at greater sizes the existence of the supersolid will persist .
the final graph shows the lr lr system . in this system
, superfluidity appears even before @xmath29 reaches @xmath66 . in this system ,
the superfluid fraction is much greater than in the previous two because of the long - ranged tunneling .
this means that at small system sizes the supersolid region is much more prominent relative to the crystal lobe .
the structure factor diminishes with system size almost exactly as in the lr - sr case except that the transition is at a different value of @xmath7 , and near @xmath63 it drops to slightly lower values .
due to this strong decrease , for any situation with long - range interactions we can not clearly state whether the supersolid region survives at larger system sizes . for sr
sr ( left top ) , lr sr ( left bottom ) , lr lr ( right ) for a @xmath62 , and @xmath42 lattice ( lines become thicker and darker with increasing system size ) .
solid blue : structure factor . dashed black : superfluid fraction . at this value of @xmath7 , for short - range tunneling
, the superfluid fraction disappears rapidly with increasing system size , while for long - range tunneling it even increases . at @xmath67
, possibly a supersolid may survive in large lattices .
again in the lr lr system there is a reversal of finite size effects .
the superfluid fraction for larger systems becomes higher instead of lower .
, scaledwidth=100.0% ] next we take vertical cuts at a value of @xmath68 , since this is a reasonable place for a supersolid to exist for the lr - lr system ( @xmath69 of the tip of the lobe ) .
we compare all the systems at this cut , and study the behavior of the superfluid fraction and the structure factor , plotted in fig .
[ vertcut15 ] . for sr
sr interactions , no supersolid region appears .
on one side of the lobe there is a sharp phase transition directly from the crystal to the superfluid phase , while on the other side there is a slower change from one solid form to another ( @xmath70 ) .
the finite - size scaling in the figure shows that as the size increases the transitions of the structure factor become even sharper , although they stay continuous due to the finite temperature .
the values of the superfluid fraction decrease as the system sizes increases and essentially disappear at @xmath34 . in the lr
sr system , a hint of the supersolid phase begins to appear on either side of the lobe .
it is a bit more evident on the side where @xmath29 is small ( as is to be expected from references such as @xcite ) , but it also arises on the opposite side .
this is contrary to a system of only short - ranged interactions where this supersolid region appears only on one side of the lobe and not both . at larger sizes , also in the lr sr system
the transitions become sharper and the superfluidity gets smaller .
the final and most interesting cut is taken out of the lr lr lobe . in this system , we see a smooth transition from crystal to supersolid at @xmath71 and at @xmath72 for @xmath33 . for larger systems
the transition at @xmath71 occurs at the same spot but becomes sharper , making the supersolid region disappear . at @xmath72
the transition shifts to a higher value of @xmath29 , making the @xmath19-filling plateau smaller .
it also becomes less smooth but the supersolid region remains longer than for the sr sr case . in the lr lr system , the superfluid fraction for larger systems has the opposite effect than for the previous cases , it becomes higher instead of lower . and
lr sr ( right ) at @xmath73 .
solid blue : structure factor . dashed black : superfluid fraction .
lines become thicker and darker as system size goes up . in the sr
sr system , a supersolid at small @xmath29 persists at large systems , while at @xmath74 the transition from crystal to superfluid becomes a direct first - order transition . for
the lr sr system , the curves change slowly at both sides of the crystal lobe , leading to persisting supersolids . , title="fig:",scaledwidth=49.0% ] and lr sr ( right ) at @xmath73 .
solid blue : structure factor . dashed black : superfluid fraction .
lines become thicker and darker as system size goes up . in the sr
sr system , a supersolid at small @xmath29 persists at large systems , while at @xmath74 the transition from crystal to superfluid becomes a direct first - order transition . for
the lr sr system , the curves change slowly at both sides of the crystal lobe , leading to persisting supersolids .
, title="fig:",scaledwidth=49.0% ] a perhaps fairer comparison is to look at a cut through a region where we are sure the supersolid exists for all three systems .
therefore , in fig . [ vertss ]
we look at two more cuts that are now taken closer to the tips of the sr sr and lr sr lobes . as with the lr - lr system ( last panel of fig .
[ vertcut15 ] ) , these lie at around @xmath75 of the tip of the lobe . in the sr sr lobe ,
the cut is taken at @xmath76 .
here we see a similar behavior for the structure factor as we did in the @xmath77 cut of the lobe , but this time the superfluid fraction plays a much more important role . on the one side , @xmath71 ,
both the superfluid fraction as well as the structure factor have sharp transitions that become even sharper at larger sizes .
in fact , at @xmath54 these transitions have been shown to be of first order , and the system goes directly from crystal to superfluid .
due to the finite temperature , they are continuous in our case . on the other side , where @xmath78 , there appears a second order phase transition into a supersolid region that spans all the way to @xmath79 . finally , we take a cut at @xmath73 of the lr sr lobe .
the behavior of this system seems to be quite different .
the first thing to notice is that the transitions on either side of the lobe are of second order . the other , and more important ,
observation is that now it appears that this system has supersolid behavior on both sides of the lobe : in addition to the expected supersolid at smaller @xmath29 , a region at @xmath29 above the crystal lobe appears where both structure factor and superfluid fraction are finite . if we recall fig .
[ vertcut15 ] , right panel , the lr lr system showed that at @xmath77 as @xmath80 increased the supersolid region disappeared from the upper side of the lobe . in the case of the lr sr system for @xmath81 the increase in the system size does not get rid of this supersolid phase .
, middle column yields lr sr at @xmath73 , right column is for lr lr at @xmath68 .
for all three systems , the three distinct quantum phases crystal , supersolid , and superfluid survive over some temperature range before they melt .
, scaledwidth=100.0% ] as a final calculation , we take a look at the important role that the temperature plays in both the melting of the crystal as well as the supersolid region . in this section , we will use the same cuts as in the previous section ( @xmath82 for sr
sr , @xmath73 for lr sr and @xmath68 for lr - lr ) so that each system will posses all the possible phases : crystal , superfluid , and supersolid .
each cut is investigated for @xmath83 at a system size of @xmath33 .
first , we analyze the structure factor to study how the @xmath31 crystal melts with an increase in temperature ( second row of fig .
[ ssmelt ] ) . for the sr
sr interactions , even at a temperature of @xmath84 there still exists a bump in the structure factor which indicates that the crystal has not completely melted yet , while for both of the long - ranged lobes the crystal melts by @xmath85 .
interestingly , the sr sr crystal and the lr lr crystal are approximately the same size at @xmath38 , but by @xmath86 one has melted while the other still exists .
that means that the system with short - ranged interactions holds its crystal structure better at higher temperatures than does our system with all long - ranged interactions .
looking at the lr sr lobe , we see that its crystal at this cut starts off smaller , yet it melts at about the same temperature as the one for lr lr interactions .
it seems that the dipolar repulsion helps stabilize the crystal structure over a larger temperature range , while the long - ranged hopping destroys the crystal more quickly because of the extra kinetic energy .
maybe more importantly , we now study the melting of the supersolid for these same cuts .
figure [ ssmelt ] shows the structure factor , superfluidity , and supersolidity as a function of temperature for each of the different systems .
since the supersolid is defined by having both non - zero structure factor and non - zero superfluid fraction , by combining the graphs we are able to see where these regions exist and also how they melt with increased temperature ( the bottom row of fig . [ ssmelt ] shows a product of structure factor and superfluid fraction , which remains finite only where the two coexist ) . a common feature of all the graphs are the spikes on either side of the plateaus .
these are regions where a phase transition occurs but does not necessarily imply that a supersolid region exists .
most likely , these features appear due to the finite size of the system and the resulting smooth transitions of superfluid fraction and structure factor . at larger sizes , the transitions would be much sharper at these points , the regions where a finite structure factor and superfluid fraction coexist would shrink , and the spikes would diminish . from the previous section , we can assume that for the sr
sr system they would disappear completely at the upper transition from the crystal lobe while for the other two systems there would still exist a small supersolid region . returning to the main focus , the small-@xmath2 region , we see that in each case a supersolid region appears that extends from the left side of the plateau all the way to @xmath63 . in every system
, this supersolid region also exists for a finite range of temperatures . for sr
sr interactions , it gradually decreases but still extends all the way out past @xmath86 . for the lr sr interactions , the supersolid region again slowly melts but now disappears at @xmath87 , just below the spot where the crystal melted .
the supersolid region for the lr lr system appears to have the largest magnitude of the three systems , but then rapidly melts at @xmath85 . , lr sr at @xmath81 and lr lr at @xmath68 ( left to right ) .
all cuts are taken at @xmath63 .
the structure factor ( solid blue ) attains similar values for all three systems .
the superfluid fraction ( dashed black ) is largest in the lr lr system and melts fastest in the lr
sr one.,title="fig:",scaledwidth=32.0% ] , lr sr at @xmath81 and lr
lr at @xmath68 ( left to right ) .
all cuts are taken at @xmath63 .
the structure factor ( solid blue ) attains similar values for all three systems .
the superfluid fraction ( dashed black ) is largest in the lr lr system and melts fastest in the lr
sr one.,title="fig:",scaledwidth=32.0% ] , lr sr at @xmath81 and lr lr at @xmath68 ( left to right ) .
all cuts are taken at @xmath63 .
the structure factor ( solid blue ) attains similar values for all three systems .
the superfluid fraction ( dashed black ) is largest in the lr lr system and melts fastest in the lr
sr one.,title="fig:",scaledwidth=32.0% ] in order to compare these transitions more quantitatively , we take a cut along @xmath63 for each system , shown in fig .
[ mumelt ] .
all three systems show a relatively similar and steady value for the structure factor .
hence , the values of the superfluid fraction are going to determine the existence of the supersolid regions .
the first plot shows the sr
sr system at the @xmath82 cut , and we can see that the superfluid fraction stays non - zero all the way out to @xmath88 .
the lr sr system has a very similar behavior at the @xmath73 cut , but in this case the supersolid is nearly completely melted by @xmath88 .
the final plot is the lr lr system at @xmath68 , which behaves slightly differently .
the most important difference is that the starting value of the superfluid fraction is higher than in the first two plots .
this should therefore make the supersolid region more pronounced .
but even though the superfluid fraction has the highest value for this system , it decays more quickly and reaches values similar to the sr
sr system at @xmath89 .
in this paper , we have presented a quantum monte carlo study of dipolar spin models that describe various systems of ultracold atoms , molecules , and ions .
we have presented predictions concerning the phase diagram of the considered systems at zero and finite temperatures , and described the appearance and some properties of the superfluid , supersolid , and crystalline phases .
while the results are not surprising and resemble earlier obtained results for similar systems in 1d and 2d , the main advantage of our study is that it is directly relevant to the current experiments : * the results for the xxz model with short - range tunneling apply for ultracold gases of polar bosonic molecules in the limit of hard - core bosons .
note that the earlier works @xcite have concentrated on the appearance of the supersolid phase and devil s staircase of crystalline phases in the square lattice @xcite , and supersolid and emulsion phases in the triangular lattice @xcite . here
we focus on the hard - core - spin limit , and compare it and stress differences with other models , such as the ones with long - range tunneling , i.e. , long - range xx interactions . *
the results for the xxz models with long - range tunneling apply for systems of trapped ions in triangular lattices of microtraps .
these results are novel , since so far such models have been only studied using various techniques in 1d , and using the mean - field approach in 2d . while the first experimental demonstrations of such models were restricted to a few ions
( see for instance @xcite ) , many experimental groups are working on an extension of such ionic quantum simulators to systems of many ions in microtraps @xcite .
in fact , very recently the nist group has engineered 2d ising interactions in a trapped - ion quantum simulator with hundreds of spins @xcite .
although in this experiment the quantum regime has not yet been achieved , it clearly opens the way toward quantum simulators of spin models with long - range interactions .
we expect that in the near future the result of our present theoretical study will become directly relevant for experiments .
this work was supported by the international phd projects programme of the foundation for polish science within the european regional development fund of the european union , agreement no . mpd/2009/6 .
we acknowledge financial support from spanish government grants toqata ( fis2008 - 01236 ) and consolider ingenio qoit ( cds2006 - 0019 ) , eu ip aqute , eu strep namequam , erc advanced grant quagatua , catalunyacaixa , alexander von humboldt foundation and hamburg theory award .
m.m . and j.z .
thank lluis torner , susana horvth and all icfo personnel for hospitality .
j.z . acknowledges support from polish national center for science project no .
dec-2012/04/a / st2/00088 .
16 b. bauer , l. d. carr , h. g. evertz , a. feiguin , j. freire , s. fuchs , l. gamper , j. gukelberger , e. gull , s. guertler , a. hehn , r. igarashi , s. v. isakov , d. koop , p. n. ma , p. mates , h. matsuo , o. parcollet , g. pawowski , j. d. picon , l. pollet , e. santos , v. w. scarola , u. schollwck , c. silva , b. surer , s. todo , s. trebst , m. troyer , m. l. wall , p. werner and s. wessel , _ j. stat
_ p05001 ( 2011 ) . | we use a quantum monte carlo method to investigate various classes of 2d spin models with long - range interactions at low temperatures . in particular , we study a dipolar xxz model with @xmath0 symmetry that appears as a hard - core boson limit of an extended hubbard model describing polarized dipolar atoms or molecules in an optical lattice .
tunneling , in such a model , is short - range , whereas density - density couplings decay with distance following a cubic power law .
we investigate also an xxz model with long - range couplings of all three spin components - such a model describes a system of ultracold ions in a lattice of microtraps .
we describe an approximate phase diagram for such systems at zero and at finite temperature , and compare their properties .
in particular , we compare the extent of crystalline , superfluid , and supersolid phases .
our predictions apply directly to current experiments with mesoscopic numbers of polar molecules and trapped ions . |
it was recently reported that over one year in a small los angeles business district , cars cruising for parking burned 47,000 gallons of gasoline and produced 730 tons of carbon dioxide @xcite . meanwhile ,
the consulting firm mckinsey recently claimed that the average car owner in paris spends four years of his or her life searching for a parking space @xcite .
the parking assignment problem associated with electric vehicles becomes even more acute .
due to the limited range of these vehicles , the marginal cost of expending energy to search for spaces may , in some cities , be prohibitively high .
thus , there is a real and compelling societal and economic need to revisit parking .
increases in car ownership , inadequate public and private parking facilities , and synchronised demand , have led to serious temporal mismatches in parking space supply and demand . finding a parking space at certain times of the day is a non - trivial challenge .
furthermore , finding spaces is not only costly in terms of wasted time for the driver , but it also creates congestion and pollution affecting everyone .
thus , improving parking improves both economic efficiency and environmental quality .
indeed , already major companies and cities are responding to these challenges .
for example , many city authorities , in order to safeguard resident on - street parking from people commuting to a city through the day , have initiated schemes such as resident permit parking , and a number of commercial initiatives have emerged in the parking area .
sfpark and justpark provide examples of city authorities and companies investing heavily in parking research and products within a smart cities context . in this work ,
our objective is not to revisit prior work on the topic .
rather , we propose a new solution that is widely applicable in cities .
we consider two nearby entities that have complementary supply and demand , i.e. , there is a shortfall of parking space at one and a surplus at the other .
_ we call the entity with a shortfall of parking a _ mini - city_. many examples come to mind : university campuses , technology parks , shopping centers , and government facilities located in the suburbs . in our work ,
we use the university campus as an example .
the second entity is the area surrounding the mini - city . during business hours , while parking spaces are limited in the mini - city , many nearby residents leave home for work . thus , there is an opportunity for the mini - city to use these vacated parking spaces , as well as from schools , hotels , apartment complexes and so on .
consequently , the mini - city has access to two classes of parking spaces : _ premium _ spaces , e.g. , those located on the university campus ; and _ secondary _ spaces located nearby and perhaps serviced by a shuttle .
note that the secondary spaces are not owned by the mini - city but rather are rented from secondary parties or _ landlords_. given this basic scheme , we consider two specific design issues .
first , we wish to guarantee a quality of service for the landlords by setting aside reserve spaces in the mini - city as contingency for events where secondary spaces are suddenly unavailable .
second , we wish to ensure that the premium spaces are allocated optimally among _ users _ ( drivers ) while preserving each user s privacy .
within the research community , questions concerning how to manage parking space supply - demand mismatch are actively being investigated from a variety of different angles .
one important aspect concerns the delivery of up - to - date , accurate , real - time information to parking systems in order to achieve greatest system efficiency .
a car park monitoring and vacancy detection system is discussed in @xcite and consists of a camera and image analysis system that reports statistics .
the aim for such systems is to provide timely information on available parking spaces and minimize search time .
parkya is an early - stage company offering a service that signposts parking locations and allows users to pay for their parking through a smartphone application .
driving directions to the parking spot are additionally available . in the near future
, such services should benefit greatly from incorporating increased amounts of real - time data , concerning up - to - date details on current parking vacancies , and predicted travel times given current road conditions .
the system of @xcite monitors and reports the state of availability of parking spaces and additionally uses context information generated by the city and its citizens to provide accurate responses to demand .
a predictive approach is proposed in @xcite and further studied in @xcite .
the authors develop a method in which car parks are able to count the number of arriving and departing cars , and report these to participating cars .
this allows them to predict the likelihood of a parking space being available at the estimated time that the car will arrive there .
this work uses ideas from queueing theory to predict the occupancy upon arrival , with car parks being modelled as single server queues with a poisson arrival process and exponentially distributed service times .
it should be mentioned that this significant reduction in requirements by using a stochastic approach comes at the cost of certainty for the customers .
the lack of a reservation system makes it possible that customers arrive to a fully occupied car park .
the main drawback of the approach is , however , that ultimately the customers will want to use the information to make a decision whether to try their luck and drive to the car park or to go somewhere else .
accordingly , there is feedback embedded in the system which needs to be taken into consideration ; namely , when users choose to drive to a car park based on the predictions made , they then affect the arrival process , rendering the model and predictions no longer valid .
this latter issue is addressed in @xcite , where multiple car park load balancing is achieved using randomisation based on congestion signals from the participating car parks .
significant extensions of this work are given in @xcite .
certainly , access to real - time information , on its own , is not enough . given the delay between an user identifying and driving to a parking space , and the fact that other users may also compete for the same parking space , a successful parking system would benefit from either intelligent assignment or reservation capability to reduce localized congestion @xcite .
the reservation system is championed in @xcite , where an architecture for an intelligent parking assistant is proposed as a public car park management solution . in @xcite ,
the parking problem is viewed as a dynamic resource - allocation problem .
similarities to problems in communication networks are drawn , for which a host of tools and methods have been developed over the last decades .
an online reservation system is proposed , where cars communicate their parking requirements and are assigned a parking space , which is then reserved and can not be used by any other vehicle . a similar approach is proposed in @xcite , albeit with a different assignment routine , and also allows the user to specify the price that he is willing to pay .
the main focus of the paper is revenue maximisation , but it is also possible to achieve other goals , such as reducing traffic levels or ensuring some sort of fairness between users from different social classes . in @xcite , an agent - based simulation model for parking facilities management was developed , the goal of the tool being to provide better understanding of how complex urban traffic systems evolve under different parking pricing strategies and different levels of parking enforcement .
agents included drivers , parking authorities , law enforcement and city government .
models for cruising - for - parking behaviour were developed in @xcite . in @xcite ,
the quality of service ( qos ) metrics is used to dimension a shared vehicle fleet such that satisfactory levels of access to the fleet were ensured to its pool of users . while the topic of @xcite was not parking per se , the paper inspires some of the notions of our current work ; that is , we consider the dimensioning of a special subset of parking spaces in our mini - city that we will call the _ reserve_. we discuss this problem in the following sections of the paper .
we will consider the following scenario .
imagine a university campus with a total of @xmath0 parking spaces , surrounded by a total of @xmath1 private parking spaces ( e.g. , from residential complexes ) , as illustrated in figure [ problem_setup ] .
a typical working day sees the university parking fill to capacity with vehicles belonging to students and staff , to the extent that the @xmath0 spaces can not meet the demand .
some campus arrivals thus have to search elsewhere . at the same time
, many nearby residents drive to work during the day and vacate their @xmath1 private parking spaces .
what we have is a wasted resource in one area ( i.e. , residential parking spaces ) and a stressed resource nearby ( i.e. , the on - campus parking ) .
we assume that a contract exists between the university and the _ landlords _ or owners of these @xmath1 private parking spaces .
the contract stipulates that these landlords lease their driveways to the university during daytime in the sequel , we will denote the time period in which the extra spaces are leased by @xmath2 $ ] .
this can be possible because the landlords drive to work during this time period . in our nomenclature ,
the premium resources are the on - campus parking spaces , while the secondary resources are the contracted residential parking spaces .
we will now consider how to design a parking system , bearing in mind the needs of the mini - city community , and the needs of the landlords .
we shall consider the following facets of the design problem : 1 .
how do we accommodate the needs of parking space owners ( landlords ) given that situations may arise when landlords will need to return home during the contracted interval ?
an example is scheduling a repairman to visit the home to fix a fitting .
( see section [ maths ] . ) 2 .
how do we accommodate the needs of the landlords given that situations arise when misbehaving university members will not vacate the parking space outside of the contracted interval ?
( see section [ maths ] . )
how do we allocate efficiently access to the premium spaces to the university community in a manner that preserves the privacy of individuals in the community ?
( see section [ fair ] . ) as we shall see , we answer these questions in a stochastic framework by developing suitable quality of service ( qos ) metrics to dimension the aforementioned parking system . to address items 1 and 2 , we shall set aside a subset of @xmath3 premium spaces as a reserve .
the research question in this latter context then becomes how to use the qos metrics to dimension the size of @xmath3 .
we shall address item 3 by applying recent ideas from distributed resource allocation .
* comment : ( generalization ) * the parking scenario presented in this paper is very specific , but our method can also be applied in the following general setting .
there are two resources , @xmath4 and @xmath5 .
these resources are atomic ( indivisible ) , with each atom being allocated to one user .
each user has a preference over the two resources . using resources @xmath4 and @xmath5 decreases a certain qos metric . in the following main sections , we answer two questions : * given a qos requirement ,
how much of resource @xmath4 and resource @xmath5 should we use ? * given an efficiency metric derived from the user preferences ,
how do we allocate the resources optimally among users in a setting where the allocation is repeated over multiple iterations ?
in this section , we consider the problem of dimensioning @xmath6 premium spaces as _ reserve _ in order to provide sufficient quality of service ( qos ) guarantees to the landlords of the secondary spaces .
let @xmath7 index the landlords of the parking spaces .
for each such landlord @xmath8 , we define a non - negative valued random variable @xmath9 , which denotes the time at which he or she returns home and needs to get the parking space back . under normal circumstances , @xmath9 is greater than @xmath10 , but on rare occasions a landlord may choose to come home early . for simplicity ,
we assume that each parking space @xmath8 has exactly one daytime user each day . for each parking space @xmath8
, we also define a non - negative valued random variable @xmath11 , which denotes the departure time of the daytime user of the space @xmath12 . under normal circumstances , @xmath11 is less than @xmath10 , but we assume a small number of miscreants so that not all spaces are always vacated on time . for convenience , recall that @xmath2
$ ] denotes the nominal rental window for every parking space . in other words , the landlord of space
@xmath12 agrees to park only outside the interval @xmath2 $ ] , whereas the daytime user of space @xmath12 agrees to park during the interval @xmath2 $ ] only .
we define the following two events for each secondary parking space @xmath8 : @xmath13 \ } \cap \{t_i
< a_i\},\\ o_i \triangleq \{w < t_i\ } \cap \{t_i < a_i\ } .\end{aligned}\ ] ] each of these events represents an outcome where a landlord would like to use the space @xmath12 , but can not do so .
the _ home - early _ event @xmath14 is due to the landlord needing the space during the day .
the _ overstay _
event @xmath15 is due to the fact that the daytime user overstayed . for easy of presentation , we assume the following properties concerning the above random variables .
[ as:1 ] the random variables @xmath16 are independent and identically distributed .
likewise , @xmath17 are also i.i.d .
moreover , all these random variables are mutually independent .
finally , we assume that all the distributions have densities .
we begin by quantifying the probability of an event @xmath15 in terms of the probability of daytime users overstaying .
recall that the distribution of the random variable @xmath11 characterises the probability of the daytime user overstaying in space @xmath12 .
first , observe that @xmath18 next , we derive an exact expression for @xmath19 using the independence assumption ( assumption [ as:1 ] ) .
let @xmath20 denote the probability distribution of @xmath21 ; and @xmath22 denote the probability distribution of @xmath23 . under assumption [ as:1 ] , we have @xmath24 we estimate @xmath20 from data in the following section
observe that @xmath25 the claim follows by definition of the distribution @xmath26 of @xmath23 .
now we derive a formula that we can use to dimension the reserve parking space @xmath3 from the contingent of @xmath0 premium spaces . recall that there are @xmath3 reserve parking spaces ( the reserve `` buffer '' ) set aside by the university . in this section
, we consider the probability @xmath27 of the event that more than @xmath3 spaces are needed to accommodate landlords needing the reserve buffer during daytime @xmath2 $ ] : @xmath28 where @xmath29 denotes a bernoulli random variable taking value 1 when event @xmath14 occurs , and value 0 otherwise .
first , we characterise the probability of the event @xmath30 in terms of the probability distributions of @xmath23 and @xmath21 .
let @xmath22 denote the probability distribution of @xmath23 .
let @xmath20 denote the probability distribution of @xmath21 . under assumption [ as:1 ] , we have @xmath31 @xmath22 and @xmath20
are estimated from data in the next section .
let @xmath32 .
observe that @xmath33 \ } \right)\\ \mbox{(by non - negativity of $ t_1$)}\quad&= { { \mathbb p}}\left ( \{t_1 < a_1\ } \cap \ { t_1 <
w \ } \right)\\ & = { { \mathbb p}}\left ( \{t_1 < a_1\ } \cap \ { t_1 < w \ } \mid a_1 \leq w\right ) { { \mathbb p}}(a_1 \leq w)\\ & \quad+ { { \mathbb p}}\left ( \{t_1 < a_1\ } \cap \ { t_1 < w \ } \mid a_1 > w\right ) { { \mathbb p}}(a_1 > w)\\ \mbox{(by bayes ' rule)}\quad&= { { \mathbb p}}\left ( t_1 < a_1 \mid a_1 \leq w\right ) { { \mathbb p}}(a_1 \leq w ) + { { \mathbb p}}\left ( t_1 < w \right ) { { \mathbb p}}(a_1 > w)\\ & = { { \mathbb p}}\left ( t_1 < a_1 \leq w\right ) + { { \mathbb p}}\left ( t_1 < w \right ) { { \mathbb p}}(a_1 > w)\\ & = \int_{t=0}^w { { \mathbb p}}\left ( t <
a_1 \leq w\right ) df_{t}(t ) + { { \mathbb p}}\left ( t_1 < w \right ) { { \mathbb p}}(a_1 > w)\\ & = \int_{t=0}^w ( f_{a}(w ) - f_{a}(t ) ) df_{t}(t ) + f_{t}(w ) ( 1-f_{a}(w)),\end{aligned}\ ] ] which is the claim . as a corollary
, we have the following expression for the probability that setting @xmath3 reserve spaces at the university is not enough .
let @xmath32 . under assumption [ as:1 ]
, @xmath27 is a random variable entirely characterized by @xmath34 : @xmath35[cor_5 ] observe that this probability can be mitigated by increasing the parameter @xmath3 . given a qos target @xmath27 equation can be used to determine the corresponding value of @xmath3 needed to achieve it .
this is illustrated through an example in the next subsection . in the previous section
we have considered arbitrary probability distributions @xmath26 , @xmath36 in the formulae derived . in this section , we give estimates @xmath37 of these distributions using publicly available data . given samples @xmath38 from the distribution @xmath26 , the corresponding empirical distribution - estimate takes the form @xmath39}.\end{aligned}\ ] ] first , we estimate the distribution @xmath36 .
recall that @xmath11 is the random variable denoting the duration of use of the @xmath12th secondary parking space . to derive an estimate , we use data on parking space utilisation collected in the city of dublin .
each data point corresponds to the time duration of one parking event .
the histogram distribution of these durations is shown in figure [ data ] .
of course , we can obtain a better estimate of the distribution of parking usage in a university campus if we have access to more particular data .
based on figure [ data ] , in order to simulate the fact that 5% of users of secondary parking spaces overstay , we set @xmath40 for our example .
next , we estimate the distribution @xmath26 for @xmath41 .
this distribution accounts for landlords who do not leave home and who arrive home normally after working hours .
there are many reasons for landlords to return home early , or not leave home at all . for simplicity
, we estimate the frequency of such days with the number of sick days of nhs staff in england over the period from april 2009 to february 2014 .
this data is presented in figure [ data2 ] . for simplicity
, we use the average sickness absence rate to estimate the probability of @xmath42 , and assume that @xmath43 otherwise ( cf . table [ fti ] )
. .probability distribution function of @xmath9 . for simplicity ,
we assume that the function is constructed so that 4.2% of landlords remain at home due to sickness , while the other 95.8% return home exactly at time instant @xmath10 . [ cols="^,^",options="header " , ] [ fti ] finally , we use corollary [ cor_5 ] to perform a dimensioning exercise based on our data . figure [ dimensioning_example ] illustrates the probability @xmath27 that @xmath3 reserve spaces are insufficient when @xmath1 secondary spaces are contracted .
the probability @xmath27 eventually falls exponentially fast versus @xmath3 .
the dependence of @xmath27 on @xmath1 is more subdued . in other words , for a fixed value of @xmath27 , a linear increase in the number of secondary spaces only requires a logarithmic increase in the number of reserve spaces . that @xmath3 reserve spaces are insufficient when @xmath1 secondary spaces are contracted ]
in this section , we consider the point of view of the user ( i.e. a staff member or student ) who requires parking at the university .
typically , such a user would purchase a monthly or yearly parking ticket .
this ticket then provides them with the opportunity to compete for a university parking space .
however , such a ticket does not necessarily guarantee them a parking space on the university grounds .
that is , if they arrive to the university `` too late '' on any given day , the car parks on - campus may already be full .
this is because more parking tickets are sold than there are actual spaces to park .
such `` first come , first served ''
systems can be inefficient .
consider the example of a studying mother or father who , even though they paid the same amount for a monthly or yearly parking ticket as everyone else , must first always take their children to school in the morning .
they thus always arrive to the campus a little bit later than everyone else and have a greater chance of missing out on an on - campus parking space .
we are interested in a scheme that offers equality in regards to access to the premium , on - campus parking spaces for all university users over the long - term period of validity of their parking ticket . for our scheme , we suppose that the university has been able to obtain enough apartment building parking spaces such that @xmath44 is greater than or equal to the number of university users requiring a parking space on any given day .
the problem then is how to allocate the premium and secondary parking spaces to users efficiently over time .
how do we ensure that one campus user does not have to park off - campus for the majority of the time while another user is almost always offered a car space on the university grounds ?
our solution to this problem is motivated by additive - increase / multiplicative - decrease ( aimd ) optimisation algorithms @xcite .
we now consider the problem of providing efficient access to the available premium spaces at the mini - city .
these spaces refer to the @xmath45 spaces available after allocating @xmath3 of the spaces as reserve according to the previous section . for simplicity of exposition
, we shall assume that @xmath46 in this section , but the results generalize in a straightforward fashion .
we introduce and recall the following notation to describe the problem data and the variables used in our algorithm . * an integer denoting the number of available premium parking spaces . *
an integer denoting the number of users wishing to avail of premium parking spaces .
we assume that @xmath47 .
* denotes discrete time , @xmath48 . in our interpretation
this corresponds to the number of days the system is operating . for convenience ,
we assume spaces are assigned on a per - day basis , but the general principles of the algorithm do not depend on this assumption .
* this is a state variable associated with the @xmath12th user .
it takes the value @xmath49 if this user is given access to a premium parking on the @xmath50th day and zero otherwise .
* this is a average access for the @xmath12th user up to the @xmath50th day : @xmath51 it is possible to formulate the premium parking space allocation problem in several ways . for example
, one could require that that the long - term average admission to the premium parking space is equal for all users , i.e. for all @xmath52 @xmath53 this assumes that all users are equal in the desire to access the premium parking space .
here we follow a more general approach and assume that each user @xmath12 has a cost function @xmath54 \to \mathbb r$ ] . for a frequency
@xmath55 $ ] of premium space allocation , the cost @xmath56 represents the monetary inconvenience cost that user @xmath12 experiences from @xmath57 .
this function specifies the priority that this user is assigned . it could represent the amount a user is willing to pay , or it could be related to the number of passengers carried by this users , or the access that this user has to public transportation ( meaning that users with fewer possibilities for alternative transport should have prioritised access to parking spaces ) . given these individual cost functions , our aim is to design a system that achieves overall minimal cost for the group .
we formulate the optimal allocation of resources as a minimization problem : @xmath58 our proposed simple algorithm for solving the parking allocation problem can be summarized as follows .
we assume that every event each user is allocated access to the premium spaces by tossing a coin .
for example , one embodiment of this idea is to use a smart - phone application .
more specifically , each user is assigned a cost function by a government authority .
for example , this could be based on vehicle class , disability , need for childcare , etc . on every subsequent day , the application outputs a variable that determines whether the user is assigned a premium space .
access to the space is assigned to the @xmath12th user with the following probability : @xmath59 a central authority that owns the parking spaces , calculates @xmath60 based on past utilization and broadcasts this scalar to all participating vehicles .
the following example illustrates the performance of this simple algorithm .
we will assume that the functions @xmath61 are continuously differentiable and strictly convex ; so that in particular the optimal point @xmath62 satisfying the constraints is unique .
furthermore , we introduce an assumption which ensures that the optimal point @xmath63 has only positive entries .
this assumption also guarantees that the algorithm we will describe is well defined for every user .
we wish to control the access to the premium space in such a way that the average utilization for each user approaches the optimal value @xmath63 , i.e. for large @xmath50 we want to achieve @xmath64 subject to the ( loose ) capacity constraint @xmath65 .
that is , all premium spaces are occupied , on average .
further , we wish to do this in a manner that preserves the privacy of users .
that is , we do not wish to reveal @xmath66 and @xmath61 to any other users in the course of the optimization .
finally , the necessary communication between the users should be minimal so as not to create a communication overhead that would be hard to sustain in an uncertain environment where users can not be expected to participate at all times . in what follows
the mechanism for preserving privacy is to develop a distributed algorithm .
we loosely follow the ideas in @xcite , where a distributed stochastic algorithm is presented which guarantees that the average utilization variables @xmath67 converge to the optimal points @xmath68 , which is also called almost sure convergence in a stochastic context . ] .
the algorithm presented here extends the ideas of @xcite , however , because we wish to ensure in addition that the instantaneous utilization variables @xmath69 sum to @xmath0 , or at least to a value close to it .
moreover , the resource to be allocated in our setting is atomic , as opposed to arbitrarily divisible .
these differences require substantial changes to the algorithm presented in @xcite . at each time
@xmath50 , each user @xmath12 determines a probability @xmath70 and sets @xmath71 where we note that all users make this probabilistic choice independently of other users or previous decisions .
the evolution of the probabilities is governed by the equation @xmath72 note that each user @xmath12 can determine its own probability with the exclusive knowledge of its own past utilization @xmath73 and cost function @xmath61 .
no information from other users is required .
the scalar @xmath60 is a network wide constant determined by the central agency and broadcast to all users . here
, @xmath60 is chosen such that @xmath74 for all @xmath75 and all @xmath76 .
it is determined in a time - varying manner as it also influences the demand for premium spaces .
specifically , if at a certain time @xmath50 each @xmath70 is fixed then the expected utilization of the premium spaces is @xmath77 moreover , the ( random ) instantaneous utilization @xmath78 is concentrated around the expected utilization by independence and hoeffding s inequality .
if we wish to ensure optimal utilization of the premium spaces and avoid overbooking , the expected utilization should be below the number of premium parking spaces .
for instance , a standard deviation below this number , depending on the desired quality - of - service metric .
denoting this number by @xmath79 , we will therefore adjust @xmath60 so that the expectation in tracks @xmath80 . as the expectation is unknown ,
we use the observed utilization as an estimator for this . taking a simple error regulation approach we thus arrive at @xmath81 the overall system is now prescribed by the dynamics of @xmath82 given by , the dynamics of @xmath83 given by and the dynamics of @xmath84 as described in .
a proof of the convergence of the algorithm is beyond the scope of this paper .
however , we present below an example to show its efficacy . for three users.,height=240 ] , height=240 ] we simulate a population of 900 users competing for 450 premium parking spaces .
each evening , users are assigned a parking space as described above ( with the scalar @xmath60 determined using a pi control ) . for simplicity ,
users have one of three cost functions : @xmath85 , @xmath86 , and @xmath87 , all of which are strictly convex .
figure [ fig : ap ] shows the average utilisation achieved for each class of vehicle .
figure [ fig : pu ] and [ fig : occ ] shows that the average utilisation of premium spaces is concentrated around the target utilisation of 450 .
figure [ fig : gain ] shows the evolution of the controller gain and figure [ fig : con ] shows that the cost functions derivatives do actually achieve consensus asymptotically . here
we will briefly comment on the heuristic which suggests that the algorithm works in a general setting . to this end
we have to point out that a number of assumptions are required .
first , in our derivation of the update rule we will rely on an infinitesimal characterization of the optimal point .
this requires that the optimal point does not lie on the boundary of the constraint set .
thus in the following we will assume that the optimal point @xmath63 satisfies @xmath88 second , the gain parameter @xmath89 in the feedback scheme needs to be chosen in such a manner that the system is stable .
noting that by the expectation @xmath90 is a multiple of @xmath60 , we choose @xmath91 where the maximum with respect to @xmath92 is taken over the constraint set .
we then reformulate to obtain @xmath93 in the previous formulation , the bracket giving the factor of @xmath60 needs to be in the interval @xmath94 sufficiently often to ensure stable behaviour of the system and thus the desired tracking property .
the choice of @xmath89 in is made to ensure that the desired stability property holds .
this ensures convergence of the sequence @xmath60 . to argue for the convergence of @xmath95 we follow the heuristic presented in @xcite .
the optimisation problem is a constrained problem for which we have assumed that the boundary conditions @xmath96 are not active .
it is then classical to introduce the lagrange parameter @xmath97 and consider the lagrangian @xmath98 from the karush - kuhn - tucker ( kkt ) conditions ( * ? ? ?
* section 5.5.3 ) , the following necessary and sufficient condition for optimality can be obtained by setting all partial derivatives to zero . in the optimal point @xmath99 ,
@xmath100 we have @xmath101 in other words , the system is at optimality when the derivatives of the cost functions are in consensus .
this observation is at the core of the choice of the probability functions in as we now explain . assuming that the system described by , , and has an ergodicity property we obtain the long - term average for the access to the premium car park satisfies ( in the limit ) @xmath102 where @xmath103 is the steady state probability that user @xmath12 has access to the resource . given this assumption , we claim that our choice of probability functions @xmath70 is such that the equation for the steady state behaviour of our system is equivalent to the kkt condition . by each user
has a probability of access that satisfies @xmath104 where the constant of proportionality is the same for all users .
suppose now that after a transient period we have @xmath105 , the steady state long - run average which exists under our ergodicity assumption
. then we can write @xmath106 .
thus , in steady state , @xmath107 and so @xmath108 for all @xmath109 .
so in the limit we expect @xmath110 for all @xmath12 .
this means that in the limit we converge to a situation where the derivatives of the cost functions are in consensus . as we have seen from the kkt conditions the consensus conditions characterizes the optimal point of the optimization problem so that we obtain @xmath111 thus the average utilization converges to the optimal point .
several comments are merited .
* this is a private algorithm .
users do not communicate with other users , nor do they reveal any state information or cost information to the central authority .
* this is a randomized algorithm .
thus we only have @xmath112 .
we argue that small fluctuations in the occupancy level about the capacity is acceptable since some users may not use the premium spaces , even if allocated one .
furthermore , in the event of all users availing of their allocated premium spaces , extra space is always available in the secondary spaces , and in the reserved spacing .
that is , the instantaneous allocation @xmath113 may exceed the capacity @xmath0 for some time steps @xmath50 , in which case , some users will have to be re - routed to the secondary parking spaces .
* the algorithm can be modified such that the parameter @xmath84 does not update at every iteration , but once in a while ( i.e. , from day to day ) . *
the proposed allocation scheme can be applied to more general problems of the form of .
for example , the same problem arises also in assigning space in overhead bins on passenger planes , seats in trains , etc . in all of these examples , users
pay a fee to compete for access to a limited resource
. however , access to the resource is not guaranteed and in effect the resource is allocated in a first - come , first - served manner . from a societal viewpoint , customers with constraints , e.g. dropping children at school , are almost always disadvantaged by such schemes .
in this work , we have modeled two aspects of parking system design and proposed two corresponding solutions .
first , we propose to complement primary or premium parking spaces with a large number of secondary parking spaces .
these spaces are obtained through contracts for shared use that provide probabilistic qos guarantees at the expense of a relatively small number of reserve premium spaces .
secondly , we propose a distributed algorithm for repeated allocation of premium spaces , which guarantees that , over time , the users who benefit most from such spaces obtain them proportionally more frequently .
the main feature of this algorithm is to keep private all information related to the benefit derived by individual users .
the next step in our research agenda is to deploy the described system on a university campus . from an algorithmic perspective
, it remains an open problem to quantify the rate of convergence of @xmath114 and to situations where we have dynamic arrivals and departures throughout each day .
another open problem is to quantify the probability of overshooting the capacity : i.e. , the event @xmath115 .
it is also useful to consider a version of the allocation algorithm for the case where a different subset of users competes in each day .
w. elfrink , the smart - city solution .
article , published by mckinsey , october 2012 .
available at : http://www.mckinsey.com/insights/public_sector/the_smart-city_solution .
( last accessed : 21 january 2015 . )
k. blumer , h. r. halaseh , m. u. ahsan , h. dong and n. mavridis , cost - effective single - camera multi - car parking monitoring ad vacancy detection towards real - world parking statistics and real - time reporting , in t. huang , z. zeng , c. li and c. s. leung ( eds . ) , _ lecture notes in computer science : neural information processing _ , vol . 7667 , 2012 , pp .
506 - 515 , springer - verlag berlin heidelberg .
j. rico , j. sancho , b. cendn and m. camus , parking easier by using context information of a smart city , in _ proceedings of the 27th international conference on advanced information networking and applications workshops _ , barcelona ,
spain , 2013 , pp .
1380 - 1385 .
m. caliskan , d. graupner and m. mauve , decentralized discovery of free parking places , in _ proceedings of the third acm international workshop on vehicular ad hoc networks _ , los angeles , california , usa , 2006 , pp .
30 - 39 .
a. schlote , c. king , e. crisostomi and r. shorten , delay - tolerant stochastic algorithms for parking space assignment , _ ieee transactions on intelligent transportation systems _ , vol .
15 , no . 5 , 2014 , pp .
1922 - 1935 .
m. dellorco , d. teodorovi and m. zambetta , an artificial intelligence approach for parking facilities management , in _ proceedings of the fourth international symposium on uncertainty modeling and analysis _
, college park , maryland , usa , 2003 , pp . 2 - 8 .
s. kaplan and s. bekhor , exploring en - route parking type and parking - search route choice : decision making framework and survey design , in _ proceedings of the second international choice modelling conference _ ,
leeds , united kingdom , 2011 . c. king , w. griggs , f. wirth , k. quinn and r. shorten , alleviating a form of electric vehicle range anxiety through on - demand vehicle access , _ international journal of control _ , vol .
88 , no . 4 , 2015 , pp .
717 - 728 . | parking spaces are resources that can be pooled together and shared , especially when there are complementary day - time and night - time users .
we answer two design questions .
first , given a quality of service requirement , how many spaces should be set aside as contingency during day - time for night - time users ?
next , how can we replace the first - come - first - served access method by one that aims at optimal efficiency while keeping user preferences private ? |
hydrodynamics is considered now as the basic part of a spatiotemporal picture of the matter evolution in the processes of ultrarelativistic heavy ion collisions ( see recent reviews @xcite ) . to complete the description of _ a+a _ collision processes , hydrodynamics must be supplied with a generator of an initial non - equilibrated state , pre - thermal dynamics which forms the initial near locally equilibrated conditions for hydro - evolution , and prescription for particle production during the breakup of the continuous medium at the final stage of the matter expansion . as for the initial state , since it fluctuates on an event - by - event basis , monte carlo event generators are widely used to simulate it in relativistic @xmath2 collisions .
the most commonly used models of initial state are the mc - glauber ( monte carlo glauber ) @xcite , mc - kln ( monte carlo kharzeev - levin - nardi ) @xcite , epos ( parton - based gribov - regge model ) @xcite , ekrt ( perturbative qcd + saturation model ) @xcite , and ip - glasma ( impact parameter dependent glasma ) @xcite . @xcite . ]
the last model also includes non - trivial non - equilibrium dynamics of the gluon fields which , however , does not lead to a proper equilibration . to apply these models to data description
some thermalization process has to be assumed .
evidently , in order to reduce uncertainties of results obtained by means of hydrodynamical models , one needs to convert a far - from - equilibrium initial state of matter in a nucleus - nucleus collision to a close to locally equilibrated one by means of a reasonable pre - equilibrium dynamics . a relaxation time method @xcite , initially developed for the post - hydrodynamic stage ,
is modified for applications to pre - thermal dynamics . as for the post - hydrodynamic ( afterburner ) stage ,
good progress was made in understanding and modeling the breakup of hadron matter into hadron gas at the late stage of expansion .
typically , particlization , i.e. , the transition from the hadron ( or quark - gluon ) fluid to hadronic gas expansions , is described by means of the so - called cooper - frye prescription within hybrid models .
in such an approach , the sudden conversion of the fluid to particles happens at a hypersurface of hadronization or chemical freeze - out .
it has long been known that such a matching prescription has problems with the energy - momentum conservation laws when fluid is converted to particles at the hypersurface which contains non - space - like parts ( see up - to - date review @xcite and references therein ) .
these problems can be avoided by using the hydrokinetic model ( hkm ) that was proposed in ref . @xcite and further developed in refs .
@xcite ( see also ref .
@xcite ) . in basic hkm ,
particlization is considered not as a sudden , as in hybrid models , but as a realistic continuous process .
in such an approach the description of continuous particle emission from the fluid is based not on the distribution function but on the so - called escape function @xcite . in refs .
@xcite the method of escape probabilities , similar to that used in hkm formalism for post - thermal dynamics , is applied to the pre - thermal one .
it makes it possible to describe the pre - thermal evolution of the energy - momentum tensor using three free parameters : the time of the initial ( non - equilibrium ) state formation , mean relaxation time , and time of thermalization , say , 1 fm / c .
this phenomenological model allows one to provide initialization of the hydrodynamical evolution , and is a more general and simple alternative to the so - called `` anisotropic hydrodynamics '' @xcite , as discussed in detail in @xcite .
the target energy - momentum tensor of the pre - thermal dynamics of _ a+a _
collisions , which the initially created system reaches during evolution , is the tensor of viscous relativistic hydrodynamics .
the further evolution is governed by the equations of the israel - stewart framework of relativistic viscous hydrodynamics until the lowest possible temperature when the system is still close to local thermal and chemical equilibrium .
it defines the hypersurface which is often called the chemical freeze - out isotherm .
the particlization procedure starts typically at this very hypersurface .
the hypersurface contains non - space - like parts that formally enforce one to use the hkm continuous particlization procedure , to eliminate the problem with violation of the energy - momentum conservation .
however , with almost smooth ( averaged ) initial conditions , the effects originated from these parts are quite small and do not affect significantly the observables @xcite .
the three - dimensional areas of non - space - like parts are small , and for large times the hydro - velocities at the corresponding parts are large .
] moreover , in viscous hydrodynamics the effects of violation of local equilibrium , which is produced in hkm for an ideal fluid , are partially taken into account .
therefore , summarizing all that and accounting for very time consuming calculations in hkm , we will use in this paper the sudden transition from hadron matter to hadron gas , similar to the hybrid models . in future studies we plan to include the option of continuous hkm particlization for situations when undesirable contributions from non - space - like parts are fairly large : it can be for @xmath3- geometry observables and event - by - event analysis .
considering also that an important component of the full model is the pre - thermal evolution where the description is built on the same idea as the basic hkm formalism , we will call the complete model , which includes generation of the initial state , its thermalization , sudden ( with the possible option of `` continuous '' ) particlization , subsequent ultrarelativistic qmd ( urqmd ) hadronic cascade @xcite , and one- and multi- particle spectra formation , the integrated hydrokinetic model ( ihkm ) .
the paper is organized as follows .
section 2 includes subsections that describe the ingredients of the complete ihkm .
section 3 is devoted to the results and discussions , while the section 4 summarizes the physical meaning of obtained results .
as discussed in the introduction the integrated hydrokinetic model ( ihkm ) consists of five ingredients which we describe below .
though prethermal dynamics is used to initialize hydrodynamics , it itself needs the initial state of the matter at the starting time @xmath4 , say @xmath5 or 0.5 fm / c , when one can speak about energy density distribution in non equilibrated matter .
following ref .
, an appropriate scale controlling the formation of gluons with a physically well - defined energy is roughly @xmath6 , where @xmath7 and @xmath8 are dimensionless parameters , which is the variance of the gaussian weight over the color charges @xmath9 of partons .
the estimate of this time gives @xmath5 fm / c for the top energy available at the bnl relativistic heavy ion collider ( rhic ) and @xmath10 fm / c for the cern large hadron collider ( lhc ) full energy @xcite .
we will utilize @xmath5 fm / c as the default value for the lhc energy @xmath11 tev ; however , we will use also @xmath12 fm / c to analyze the corresponding variation of the results .
we attribute the initial state to the time @xmath4 .
the generation of the initial state is based on the glissando 2 @xcite package .
this package works in the frame of the semiclassical glauber model . within this approach at the very initial stage of the collision , individual interactions between nucleons deposit transverse energy .
each deposition of the transverse energy at a certain space - time point or region is called a source and each source has its weight which is called relative deposited strength , or rds .
the normalization of rds can be treated as an insertion of the additional parameter .
we choose its value in such a way that all charged - particle multiplicity distribution , obtained in ihkm for 5% most central collisions , fits its experimental value .
the rds may be different for the wounded nucleons and the binary collisions , and it can fluctuate from source to source .
we use a mixed model , amending the wounded nucleon model with some binary collisions . in this model
, a wounded nucleon obtains the rds of @xmath13 and a binary collision has the rds of @xmath14 .
the total rds , averaged over events , is then @xmath15 .
of course , the result of simulations fluctuates from event to event . in this paper
we do not study fluctuations of observables , and will simulate inclusive observables basing on the approximation of the averaged initial state for each centrality class .
this approximation implies that for each selected centrality only the one hydro run is performed .
the simulation of a single event is made in three stages : 1 .
generation of the positions of nucleons in the two colliding nuclei according to the fluctuating nuclear density distribution .
the form of the mean distribution depends on the mass of nuclei @xmath16 . for sufficiently large nuclei
this distribution has a woods - saxon form with an account of nuclear deformations , the latter are small in our case of pb+pb collisions .
generation of the transverse positions of the sources and their rdss .
3 . calculation of the physical quantities and writing the results in the output file . in order to obtain the initial conditions from the rds
, one can put the rds to be proportional to energy or entropy .
we choose it to be proportional to energy because entropy is not created yet in a non - equilibrium initial state .
then , for the initial state averaged over many events with a chosen centrality bin , one can write ( @xmath17 is the impact parameter associated with centrality ) @xmath18 where @xmath19 is a fitting parameter defining the maximal energy density of the initial state that is reached at minimal centrality , say 05% ( @xmath20 ) , at @xmath21 .
the value @xmath19 and @xmath14 are the same for all centrality classes .
the values of @xmath19 and @xmath14 are fixed by the reproduction of the charged - particle multiplicity at different centralities .
the generation of the initial energy density configuration with the help of glissando is supposed to be followed by a thermalization process . as we have already discussed in the introduction
we describe the matter evolution at the pre - thermal stage with the relaxation model @xcite .
we assume a longitudinal boost invariance , which is a good approximation for the central rapidity region of the fireball at high collision energies .
the relaxation model simulations are applied to the system for the early stage of the heavy ion collision , from the proper time @xmath22 to @xmath23 , where @xmath24 . as mentioned above
, we use @xmath5 fm / c as the basic value and @xmath12 fm / c for a comparison .
thermalization time is supposed to be @xmath25 fm / c , the``conventional '' time of the formation of strongly interaction matter , @xmath26 fm / c and with the other parameters to be the same as in our basic scenario ( see later ) lead to very similar results as calculations with @xmath25 fm / c . ]
the total energy - momentum tensor of the matter is taken in the form @xcite @xmath27 , \label{1}\end{aligned}\ ] ] where @xmath28 and @xmath29 are hydrodynamic ( local equilibrium ) and free ( fully or almost free evolving ) components of the energy - momentum tensor , @xmath30 is the weight function satisfying the following conditions : @xmath31 , @xmath32 , @xmath33 , and @xmath34 .
its form will be discussed below . as for the free ( or almost free ) evolving component
, it is well defined in such models as ip - glasma and epos , but not in mc glauber .
to analyze different kinds of anisotropy of the initial state we present the boost - invariant distribution function at the initial hypersurface @xmath35 : @xmath22 in the factorized form @xmath36 where @xmath37 is defined by eq .
( [ rds ] ) . at other times
@xmath38 is defined by the free - streaming requirement @xmath39 , so that at all times the free evolution of the energy - momentum tensor is defined by the formula @xmath40 then the evolution of @xmath29 and corresponding distribution function @xmath38 are defined by the initial conditions , which are generated by glissando , and the initial momentum anisotropy of the function @xmath41 . for @xmath32 , we have @xmath42 . the hydrodynamic component of the energy - momentum tensor has the form @xmath43u^{\mu}_{\text{hydro}}(x)u^{\nu}_{\text{hydro}}(x ) - [ p_{\text{hydro}}(x)+\pi]g^{\mu \nu } + \pi^{\mu \nu } , \label{6}\end{aligned}\ ] ] where @xmath44 is the energy density in the local rest frame , @xmath45 is the pressure , @xmath46 is the shear stress tensor , @xmath47 is the bulk pressure , and @xmath48 is the energy flow four - vector .
we neglect the bulk pressure , @xmath49 . in curvilinear ( hyperbolic ) coordinates the equation of motion for the shear stress tensor is taken in the form @xmath50 where the semicolon means the covariant derivation , brackets are defined as @xmath51 , @xmath52 , and @xmath53 is the navier - stokes shear stress tensor : @xmath54 for @xmath30 we use the ansatz @xcite @xmath55 where , as mentioned above , @xmath5 fm / c , @xmath25 fm / c , and @xmath56 .
the value of @xmath57 is the model parameter , which characterizes the rate of thermalisation , that is , the speed of conversion of the non - equilibrium state to about the equilibrated one .
writing down the conservation laws for the total energy - momentum tensor in the form @xmath58 and accounting that for the free streaming @xmath59 , we have @xmath60t^{\mu \nu}_{\text{hydro}}(x)\}= - t^{\mu \nu}_{\text{free}}(x)\partial_{;\mu}{\cal p}(\tau ) .
\label{7}\end{aligned}\ ] ] let us introduce the re - scaled hydrodynamic tensor @xmath61 with initial conditions @xmath62 at @xmath63 for all @xmath64 . then @xmath65 this is the hydrodynamic - type equation of motion with a source . as mentioned above
, @xmath66 is defined by the the initial state at proper time @xmath4 , and @xmath67 is defined explicitly , then the source in the right - hand side of eq .
( [ 8 ] ) can be calculated for all @xmath68 and @xmath64 .
the introduction of the re - scaled hydrodynamic tensor leads to the rescaling of the shear stress tensor @xmath69 .
then , multiplying eq .
( [ evolution-1 ] ) by @xmath70 , we have the equation of motion for the re - scaled shear stress tensor @xmath71\left \langle u^\gamma \partial_{;\gamma } \frac{\widetilde{\pi}^{\mu\nu}}{(1-{\cal p}(\tau))}\right \rangle = -\frac{\widetilde{\pi}^{\mu\nu}-[1-{\cal p}(\tau)]\pi_\text{ns}^{\mu\nu}}{\tau_\pi}-\frac { 4}{3 } \widetilde{\pi}^{\mu\nu}\partial_{;\gamma}u^\gamma . \label{evolution-3}\end{aligned}\ ] ] we also have to specify the equation of state in order to close the set of evolutionary equations . in all calculations
the laine - schroeder equation of state ( eos ) @xcite is applied . at @xmath72
we switch to the viscous hydrodynamic approach .
the equations of motion are nearly the same as in eq .
( [ 8 ] ) , but with the right - hand side equal to zero .
at @xmath73 fm / c the @xmath33 and the target function is reached : @xmath74 .
the further evolution is described by the relativistic viscous hydrodynamics according to the equations ( [ 6 ] ) - ( [ evolution-2 ] ) , ( [ 8 ] ) , and ( [ evolution-3 ] ) with @xmath75 .
a numerical solution of the viscous hydrodynamic equations is constructed with the ` vhlle ` code @xcite .
such an evolution describes the expansion of superdense quark - gluon and hadron matter close to local chemical and thermal equilibria with a baryon chemical potential @xmath76 ( which is a good approximation for lhc energies ) until the temperature when such an approach breaks down .
then the system has lost the properties of local equilibrium , thermal and chemical as well , and another approximation should be used . as discussed in the introduction ,
the basic hkm describes particlization as a continuous process ( for review see @xcite ) : particles gradually `` escape '' from the expanding fluid , forming the non - equilibrium wigner function , which then can be used at some space - like hypersurface as the input for urqmd hadronic cascade @xcite . as it was already discussed , at smooth initial conditions the results are very similar to a sudden particlization scenario and we will use the latter in this paper .
we assume that the chemically and thermally locally equilibrated evolution takes place until temperature @xmath77 mev ( corresponding to an energy density @xmath78 gev/@xmath79 for the laine - schroeder eos ) is reached , and switch to particle cascade at the hypersurface defined by this criterion .
such a switching surface is built during the hydrodynamic evolution with the help of the cornelius @xcite routine .
we apply the cooper - frye formula to convert the fluid to the cascade of particles : @xmath80 in order to take into account the viscous corrections to the distribution function , grad s 14-moment ansatz is used .
we assume that the corrections are the same for all hadron species
. then formula ( [ cooper - frye1 ] ) transforms to @xmath81 \label{cooper - frye2}\end{aligned}\ ] ] this distribution function is used to create an ensemble of particles on the hypersurface . at first
, the average number of hadrons of every sort is calculated : @xmath82 the average total number of particles then equals @xmath83 . the exact total number of the particles to be created is sampled according to poisson distribution with a mean value @xmath84 .
the type of each generated particle is chosen randomly based on probabilities @xmath85 .
then the momentum is assigned to each particle in the local rest frame of the fluid .
the direction of momentum is chosen randomly in the @xmath3 solid angle and its modulus is generated according to the isotropic part of eq .
( [ cooper - frye2 ] ) .
after that , the corrections for @xmath86 are applied via the rejection sampling .
the particle position is set to be equal to the position of centroid of the surface element , and the spacetime rapidity is sampled randomly within the longitudinal size of the volume element .
finally , the particle momentum is lorentz boosted to the center - of - mass frame of the fireball .
the generated hadrons are then fed into the urqmd cascade .
since the cascade accepts only a list of particles at an equal cartesian time as an input , the created particles are propagated backward in time to when the first particle was created .
the particles are not allowed to interact in the cascade until their trajectories cross the particlization hypersurface .
the laine - schroeder eos , that is applied in our analysis corresponds to an equilibrium hadron - resonance gas consisting of about 360 hadrons in the low temperature limit .
many of those heavy hadrons are not included in the urqmd hadron list . to prevent violation of the energy - momentum conservation even at space - like parts of the isotherm @xmath77 mev , we decay the heavy resonances , which are not in the urqmd list , just at the switching hypersurface .
the particle propagation is stopped at cartesian time 400 fm / c , where their coordinates and momenta are recorded .
we generate 50000 glissando events for each centrality class to produce averaged initial energy density profile for the pre - thermal / hydrodynamic evolution .
only the one hydrodynamic calculation is performed for each centrality class with a given set of parameters . for each hydrodynamic evolution
20000 urqmd events are generated .
the generated sets of events are stored in ` root ` trees and files , which are further processed with scripts to plot momentum distributions and calculate flow coefficients or correlation functions for physical analysis .
to perform calculations of the spectra , anisotropic flow coefficients , and interferometry radii we need to specify the initial states and model parameters .
as discussed above , we generate these initial conditions with glissando 2 , a monte carlo glauber generator .
the output of this generator is the relative deposited strength ( rds ) , which defines the initial energy density ( [ rds ] ) through the number of wounded nucleons @xmath87 and number of binary collisions @xmath88 .
the relative weight of these quantities @xmath14 is crucial for the description of all charged - hadron multiplicities at different centralities .
it is found that the value @xmath89 at different reasonable values of other parameters gives the best fit for the multiplicities at lhc energy @xmath90 tev .
figure [ fig : mult ] demonstrates a typical fit at this value of @xmath14 .
fm / c , @xmath91 fm / c , @xmath92 , and @xmath93 for the lhc energy @xmath94 tev .
experimental data are from @xcite.,scaledwidth=55.0% ] to analyze the effects of momentum anisotropy of the initial state , similarly to ref .
@xcite we choose the momentum dependence of the distribution function @xmath41 in eq .
( [ f0 ] ) in the form @xcite : @xmath95 where @xmath96 , @xmath97 .
note , that in the rest frame of the fluid element , @xmath98 , @xmath99 and @xmath100 .
then one can see that the parameters @xmath101 and @xmath102 can be associated with the different temperatures along the beam axis and orthogonal to it correspondingly .
the parameter @xmath103 is the main parameter , that defines momentum anisotropy of the initial state .
we use two values for it : @xmath104 ( momentum isotropic case ) and @xmath93 ( anisotropic case , almost no pressure in the longitudinal direction ) .
we put @xmath105 gev @xcite . at @xmath106
the last value is just absorbed into a factor @xmath107 that is the normalization constant in eq .
( [ anis1 ] ) .
this constant is defined from the requirement that the initial energy - momentum tensor zero component , formally associated with the dimensional function @xmath41 only and so calculated according to eq .
( [ 2 ] ) with substitution @xmath108 , is unity : @xmath109=1 $ ] [ see eqs .
( [ rds ] ) and ( [ f0 ] ) ] . as for the anisotropy factor @xmath110
, it also does not affect the initial energy density distribution ( [ rds ] ) , but reveals itself in the process of the matter evolution at the pre - thermal stage , see sec .
ii b. the generated distribution functions are related to the two initial times @xmath5 and @xmath12 fm / c for comparison . the thermalization time is chosen to be , as discussed before , @xmath111 fm / c .
when other parameters are fixed , we make calculations for 05% , 510% , 1020% , 2030% , 3040% , 4050% , and 5060% centralities with the same normalization coefficient @xmath19 in eq .
( [ rds ] ) and fixed relative contribution of binary collision @xmath89 .
the latter , as we discussed above , does not in fact depend , on all the other parameters , contrary to normalization @xmath19 .
the @xmath19 values at different scenarios , that are associated with ihkm parameters , are shown in table 1 . [ cols="^,^,^,^,^,^,^",options="header " , ] table 1 .
the maximal initial energy densities @xmath19 and the mean @xmath112 for different scenarios . the values @xmath4 , @xmath57 correspond to fm/@xmath113 .
the mean @xmath114ndf is calculated by an averaging @xmath114ndf values taken from model comparison with observed pion , soft pion , kaon , and antiproton spectra , mid - rapidity charged multiplicity dependence on centrality , @xmath115 transverse momentum dependence for charged particles , and the pion interferometry radii as functions of the transverse momentum . for each set of parameters with corresponding @xmath19 (
table 1 ) , the experimental multiplicity dependence on centrality is reproduced well , similar to fig .
[ fig : mult ] , with the same @xmath89 .
there is a correspondence between the ratios of @xmath19 at the different parameters and inverse ratios of the energy densities at the thermalization time @xmath111 fm / c , obtained at the such parameters in @xcite when the initial energy densities at @xmath4 are the same . after specifying the initial conditions we run the relaxation model , which describes the pre - thermal stage with different relaxation times @xmath91 and 0.75 fm / c .
the grid spacing and time step for the relaxation model and further pure hydrodynamic calculations are @xmath116 fm / c and @xmath117 fm / c . here
we perform 2 + 1-dimensional longitudinally boost - invariant calculations in ihkm , which is a good approximation for the central rapidity interval at lhc energies . at the thermalization ( proper )
time @xmath25 fm / c , the evolutionary equations at the pre - thermal stage are smoothly switched ( automatically ) to relativistic viscous hydrodynamic equations in the israel - stewart framework .
we use the two values of the shear viscosity to entropy ratio in the hydrodynamic phase : the minimal one @xmath118 and @xmath119 for comparison .
also we compare the basic scenario ( bs ) with @xmath5 , @xmath91 fm / c , @xmath92 , anisotropy parameter @xmath93 , which is selected as giving the optimal description of the experimental data ( see below ) , with results of calculations with other different parameters , including viscous and ideal pure hydrodynamic scenarios ( without the pre - thermal stage but with the subsequent hadronic cascade ) .
gev/_c _ region for various centrality classes obtained in the ihkm basic scenario ( as in fig .
[ fig : mult ] ) .
the results are compared with those in ihkm at the other parameter @xmath12 fm / c and with pure viscous hydro at the starting time @xmath120 fm / c for centrality classes 05% , 1020% , and 3040% .
the experimental data are from @xcite .
the spectra for different centralities are multiplied by a factor of 2 ( @xmath121 for 05% centrality ) .
the lines correspond to different centrality classes as marked by letters : a , 05% ; b , 510% ; c , 1020% ; d , 2030% , e , 3040% ; f , 4050% ; g , 5060%,scaledwidth=80.0% ] region for 05% centrality in the ihkm basic scenario ( as in fig .
[ fig : mult ] ) in comparison with the results obtained with ( 1 ) the other relaxation time @xmath122 fm / c instead of 0.25 fm / c , or with ( 2 ) the isotropic parameter @xmath104 instead of the anisotropy one @xmath93 , or with ( 3 ) the other initial time @xmath12 fm / c instead of 0.1 fm / c . also the results for pure viscous hydro , starting at @xmath120 fm / c are presented .
the experimental data are from @xcite.,scaledwidth=80.0% ] .
, scaledwidth=85.0% ] .
, scaledwidth=85.0% ] in fig .
[ fig : pi1 ] we compare the results with the pion spectra at @xmath123 gev/_c_. as one can see at fairly large @xmath124 gev/_c _ the results for pion transverse spectra in the scenario with initial time @xmath125 fm / c describe poorly the experimental data : @xmath126/ndf=14.33 .
the best plots correspond to shear viscosity value @xmath119 , @xmath126/ndf=2.48 , in the basic scenario ( bs ) with @xmath127 .
the bs itself leads to satisfactory results for pion spectra also .
according to table 1 the basic scenario leads to mean value @xmath128=3.68 over all observables under consideration .
while the scenario with @xmath127 has @xmath129= 6.61 because it does not describe @xmath115 ( see below fig .
[ fig : v2 ] ) , @xmath130/ndf=27.92 .
in addition to the basic scenario the relatively good description of the wide class of observables gives , according to table 1 , the similar scenarios with , however , larger relaxation time , @xmath91 fm / c @xmath131 fm / c , and the scenario with isotropic initial state , @xmath132 .
so , for more detailed analysis , let us consider the interesting phenomenon of soft pion enhancement presented in fig .
[ fig : pi2 ] .
as one can see , the ihkm scenarios with isotropic initial state , @xmath104 or with larger relaxation time , @xmath122 , obviously underestimate @xmath133 spectra in the soft region , namely , @xmath134/ndf=1.03 for the isotropic case and @xmath134/ndf=1.27 for @xmath135 0.75 fm / c ; pure ( viscous ) hydrodynamics scenario with @xmath136 fm / c is also less effective at soft @xmath133 for pions . at the same time , the basic scenario leads to twice less value @xmath134/ndf=0.59 .
the description of this soft momentum region becomes the best when the initial state formation is ascribed to the later time @xmath12 fm / c .
however , it excludes , as we discussed above ( see also table 1 ) , a satisfactory spectra description at @xmath137 gev/_c _ despite that there is no problem with @xmath115 coefficients and the interferometry radii .
this could mean the necessity of including mini - jets in the hydrokinetic picture in order to describe the large @xmath133 spectra simultaneously with very small transverse momenta .
one of the other proposals considers the specific mechanism of soft pion radiation , namely , the bose - einstein condensation in the thermal model @xcite .
however , since the pion spectra in the basic scenario are within the experimental errors with small @xmath134/ndf , we consider now the bs as the most realistic , allowing to describe well the bulk observables in wide @xmath133 region
. as one can see from fig .
[ fig : k1 ] , ihkm also describes well the kaon spectra in the basic scenario , @xmath138/ndf=1.9 . as for the antiproton spectra ( fig .
[ fig : aproton1 ] ) , one can see that the basic scenario is not the best one ( fig .
[ fig : aproton1 ] ) . that is why
, if one compares the mean values of @xmath128 over all observables in table 1 , this scenario brings slightly worse results than scenarios with an isotropic initial state and anisotropic one with larger relaxation time . at the same time
the last two scenarios do not describe the enhancement of soft pions as we already discussed .
let us apply ihkm to a description of elliptic flows .
the elliptic flows in ihkm are associated with the reaction plane that is well - defined in glissando 2 .
we compare the results with the experimental @xmath115-coefficients estimated with four - particle cumulants measured for unidentified charged particles as a function of transverse momentum for various centrality classes , @xmath139 @xcite .
figure [ fig : v2 ] demonstrates all charged - particle @xmath115-coefficients for the basic scenario in comparison with ihkm results with @xmath119 instead of @xmath92 and also with ideal hydro starting at @xmath136 fm / c .
one can see that the two last scenarios lead to poorly described @xmath115-coefficients . at the same time
the basic scenario , as well as the ihkm one with @xmath12 instead of @xmath5 fm / c and also the viscous pure hydrodynamics scenario with @xmath136 fm / c describe the data quite satisfactorily , see fig .
[ fig : v2_2 ] .
however , the two last scenarios describe the spectra worse : the one with @xmath12 , at fairly large @xmath133 ; the other , the pure viscous hydro , at both large and small momenta .
coefficients for centrality classes 05% , 510% , 1020% , 2030% , 3040% , 4050% and 5060% , obtained in the ihkm basic scenario ( as in fig .
[ fig : mult ] ) .
the results are compared with those in ihkm at the other parameter dissipation condition , @xmath119 instead of @xmath140 and with ideal hydro with the starting time @xmath141 fm / c for centrality classes 510% , 1020% , and 2030% .
the experimental data are from @xcite.,scaledwidth=75.0% ] , but the results are compared with those in ihkm at the other initial time , @xmath12 fm / c instead of @xmath142 fm / c and with viscous hydro at the starting time @xmath141 fm / c.,scaledwidth=75.0% ] an important observation for evolutionary models is the interferometry radii that reflect the space - time structure of the particle emission from an expanding fireball .
the detailed study within ihkm shows that if a selection of some set of parameter values , such as @xmath4 , @xmath57 , @xmath110 , @xmath143 , is accompanied by the renormalization of the maximal initial energy density @xmath19 , similar to that in table 1 , to keep the ihkm charged - particle multiplicity to be equal to the experimental one , the interferometry radii change within a few percent only , which is a much smaller change than the one caused by a change in the centrality class of the collision .
published results ( see , e.g. review @xcite ) have shown that the interferometry radii depend on particle species , multiplicity , and initial system size .
the other details are not so important .
dependence on transverse momentum for different centralities in the ihkm basic scenario under the same conditions as in fig .
[ fig : mult].,scaledwidth=80.0% ] dependence on transverse momentum for different centralities in the ihkm basic scenario under the same conditions as in fig .
[ fig : mult ] .
the experimental data are from @xcite.,scaledwidth=80.0% ] dependence on transverse momentum for different centralities in the ihkm basic scenario under the same conditions as in fig .
[ fig : mult ] .
the experimental data are from @xcite.,scaledwidth=80.0% ] ratio for different centralities in the ihkm basic scenario under the same conditions as in fig .
[ fig : mult ] .
the experimental data are from @xcite.,scaledwidth=80.0% ] in figs .
[ fig : rout][fig : rlong ] we present the pion interferometry radii in the ihkm basic scenario for three different centralities .
in general the radii are described well except some small deviation for @xmath144 and @xmath145 from the central experimental points ; these deviations are oppositely directed for these transverse radii , which results in slight overestimation of the @xmath145 to @xmath144 ratio , as demonstrated in fig .
[ fig : rors ] .
the integrated hydrokinetic model ( ihkm ) of _ a+a _ collisions is developed .
it includes the generation of the initial ( generally momentum anisotropic ) state , matter evolution at the pre - thermal stage leading to thermalization , subsequent viscous hydrodynamic expansion , particlization , and hadronic cascade urqmd .
this model is applied to describe at different centralities the mid - rapidity charged - particle multiplicities , the pion , kaon , and antiproton spectra , the @xmath115 coefficients for all charged hadrons , and the pion interferometry radii in pb - pb collisons at lhc with @xmath94 tev .
it is found that a quite satisfactory description of the bulk observables can be reached when the initial states are attributed to the very small initial time 0.1 fm / c , the pre - thermal stage ( thermalization process ) lasts at least until 1 fm / c , and the shear viscosity at the hydrodynamic stage of the matter evolution has its minimal value , @xmath1 .
the requirement of satisfactory description of the soft pion enhancement phenomenon also leads to the further discrimination between the ihkm scenarios , namely , the initial state should be maximally momentum anisotropic and have a small mean relaxation time , @xmath146 . in this basic scenario
the bulk observables are described in a wide @xmath133 region , including the pion and antiproton spectra , as well as their yields when annihilation processes at the afterburner ( urqmd ) stage are taken into account @xcite .
it is worthy noting that while the found tendencies , which direction the changes of spectra and @xmath115 values go when changing the basic model parameters , are probably stable , an accounting of event - by - event fluctuations , as well as modification of such important factors as eos and particlization temperature , can change the numerical values of the model parameters , ensuring an optimal description of the observables .
such an investigation is the subject of subsequent work .
another important point which we would like to stress is the approximate similarity of the results at different reasonable values of the main parameters of the model , even including the pure hydrodynamic scenario starting from @xmath147 fm / c .
the reason for such similarity is probably that each change of basic parameters is accompanied by a re - normalization of the maximal energy density of the initial state , to keep the agreement with the experimental multiplicity of all charged particles at midrapidity .
then , at the same particlization temperature for all the scenarios , the main characteristics of the bulk observables are approximately preserved .
of course , some details are different , as ihkm demonstrates , but the variations are not so dramatic .
this observation explains satisfactory agreement of different variants of hydrodynamic and hybrid models with the experimental data for a+a collisions , especially when not all bulk observables are considered . in this paper the monte carlo glauber model
is used for the generation of the initial state . the profile from glissando
is associated here with the one for energy density of the non - equilibrium and non - thermal initial state .
note that the entropy density can not be used to characterize such a state at @xmath5 fm / c .
the interesting point is that with such treatment of the monte carlo glauber model , the share of the binary collisions , contributed to the total energy density , is constant at about 1/4 for lhc energies at _ different _ ihkm parameters .
we are planning to utilize other initial state models , such as epos and ip - glasma , which have other structures of the initial state and already fixed values of the maximal initial energy density , as the generators of the initial states in ihkm .
the comparison of the models have to be done on an event - by - event basis and include besides spectra , their anisotropy and femtoscopic scales for different hadron pairs , as well as fluctuations of the observables .
the research was carried out within the scope of the eurea : european ultra relativistic energies agreement ( european research group : `` heavy ions at ultrarelativistic energies '' ) , and is supported by the national academy of sciences of ukraine ( agreements f7 - 2016 ) .
i.k . acknowledges financial support by the helmholtz international center for fair and the hessian loewe initiative .
u. heinz , j. phys .
. ser . * 455 * , 012044 ( 2013 ) ; u. heinz , and r. snellings , annu .
nucl . part .
sci . * 63 * , 123 ( 2013 ) [ arxiv:1301.2826 ] ; c. gale , s. jeon , b. schenke , int .
a * 28 * , 1340011 ( 2013 ) ; p.
huovinen , int .
e * 22 * , 1330029 ( 2013 ) [ arxiv:1311.1849 ] ; h. song , arxiv:1401.0079 [ nucl - th ] ( 2014 ) m. alvioli , h .- j .
drescher , m. strikman , phys .
b * 680 * , 225 ( 2009 ) ; h. holopainen , h. niemi , k. j. eskola , phys . rev .
c * 83 * , 034901 ( 2011 ) ; m. alvioli , h. holopainen , k. j. eskola , m. strikman , phys . rev .
c * 85 * , 034902 ( 2012 ) .
d. kharzeev , c. lourenco , m. nardi , h. satz , z. phys .
c * 74 * , 307 ( 1997 ) ; d. kharzeev , m. nardi , phys . lett .
b * 507 * , 121 ( 2001 ) ; d. kharzeev , e. levin , phys .
b * 523 * , 79 ( 2001 ) ; d. kharzeev , e. levin , m. nardi , phys .
c * 71 * , 054903 ( 2005 ) .
akkelin , y. hama , iu.a .
karpenko , yu.m .
sinyukov , phys .
c * 78 * , 034906 ( 2008 ) ; yu.m .
sinyukov , s.v .
akkelin , iu.a .
karpenko , y. hama , acta phys .
b * 40 * , 1025 ( 2009 ) [ arxiv:0901.1576 ] ; yu.a .
karpenko , yu.m .
sinyukov , j. phys .
g * 38 * , 124059 ( 2011 ) [ arxiv:1107.3745 ] ; | a further development of the evolutionary picture of _ a+a
_ collisions , which we call the integrated hydrokinetic model ( ihkm ) , is proposed .
the model comprises a generator of the initial state glissando , pre - thermal dynamics of _ a+a _
collisions leading to thermalization , subsequent relativistic viscous hydrodynamic expansion of quark - gluon and hadron medium ( vhlle ) , its particlization , and finally hadronic cascade ultrarelativistic qmd .
we calculate mid - rapidity charged - particle multiplicities , pion , kaon , and antiproton spectra , charged - particle elliptic flows , and pion interferometry radii for pb+pb collisions at the energies available at the cern large hadron collider , @xmath0 tev , at different centralities .
we find that the best description of the experimental data is reached when the initial states are attributed to the very small initial time 0.1 fm / c , the pre - thermal stage ( thermalization process ) lasts at least until 1 fm / c , and the shear viscosity at the hydrodynamic stage of the matter evolution has its minimal value , @xmath1 . at the same time
it is observed that the various momentum anisotropies of the initial states , different initial and relaxation times , as well as even a treatment of the pre - thermal stage within just viscous or ideal hydrodynamic approach , leads sometimes to worse but nevertheless similar results , _ if _ the normalization of maximal initial energy density in most central events is adjusted to reproduce the final hadron multiplicity in each scenario .
this can explain a good enough data description in numerous variants of hybrid models without a prethermal stage when the initial energy densities are defined up to a common factor . |
in the current paradigm for cosmological structure formation , dark halos collapse from initial gaussian density fluctuations and grow by accretion and merging in a hierarchical fashion .
a longstanding prediction of the theory is that the subhalo ( or satellite ) population is self - similar , meaning simply that low mass systems , such as galaxies are scale - down versions of larger systems , like galaxy clusters . a galaxy such as the milky way
is predicted to have nearly the same scaled distribution of substructures as a more massive galaxy cluster such as virgo ( moore et al .
1999 , klypin et al .
this prediction is tested by using the substructure velocity distribution function that expresses the number of sub - halos with circular rotational velocity @xmath9 greater than a certain fraction of the circular velocity of the parent halo @xmath10 .
cosmological numerical simulations confirm the expectations predicting about 500 satellites with @xmath11 within 500 kpc of the milky way .
accounting for the most recent discoveries of sloan data for fainter galaxies ( i.e. willman et al .
2005 , kleyna et al . 2005 ) up to date less than 30 satellite galaxies have been observed in the local group including underluminous systems that likely have even lower @xmath12 pointing out a missing satellite problem in the current paradigm for structure formation .
proposed solutions to this puzzle invoke the reionization of the universe at high redshift combined with the local injection of energy from star formation ( bullock et al .
2000 , somerville 2002 , benson et al .
these mechanisms may be invoked to suppress galaxy formation in halos below @xmath13 reconciling the predictions with observations in the local group .
recently , however , donghia & lake ( 2004 ) have shown that fossil groups with intermediate mass between the local group and the virgo cluster are systems with paucity of substructures pointing to a missing galaxy problem in higher mass systems . in fossil groups , a giant e or cd galaxy dominates with the next brightest group member being at least 2 r - band magnitudes fainter .
the archetype is rxj1340.6 + 4018 , an isolated , bright ( m@xmath14 ) elliptical galaxy at z=0.171 , surrounded by a halo of hot gas and dark matter ( jones et al .
its total mass is @xmath15% of the virgo cluster . with @xmath16% of virgo ,
the escape energies of their missing substructures in the fossil group are @xmath17 times larger than the satellites missing from the local group , far too large to be explained by proposed suppression mechanisms .
subsequent observations have shown that fossil groups are clearly virialized objects in which a central galaxy is embedded in a x - ray halo with a luminosity 10 - 60 hickson groups , many of which show only a faint x - ray emission and may be either evolving systems ( miles et al .
2004 ) or fake associations ; they are more useful than loose groups , which are often unvirialized structures contaminated by sighting down filaments ( hernquist et al .
1995 ) .
recently , however , kravtsov et al . (
2004 ) proposed that luminous dwarf spheroidals in the local group descend from downsizing of relatively massive ( @xmath18m@xmath19 ) high - z systems by processes such as tidal stripping .
again , this is idea may help explaining the lack of substructures in the local group ( mayer et al .
2007 ) but it is unlikely to explain how massive objects ( @xmath20 m@xmath19 ) survive in a virgo - size object where the ram stripping is more efficient and yet are absent in fossil groups of nearly the same mass . is rxj1340.6 + 4018 representative of the whole class of fossil groups ?
early work on the fossil groups focused on the magnitude gap in the luminosity function and suggested that it owed to early formation times and subsequent merging ( jones et al . 2003 , donghia et al . 2005 )
. however , little work has been devoted to understand the truly remarkable feature that nearly every group with a magnitude gap of 2 has a substructure function like the local group .
the known exceptions are rx j1552.2 + 2013 ( e.g. cypriano , mendes de oliveira and sodr 2006 ) and a cluster of galaxies discovered recently ( gastaldello et al .
2007 ) which are rich enough with substructure that they have been called `` fossil clusters '' .
the self - similarity of the subhalo population resulting from cosmological n - body simulations is still debated ( e.g. de lucia et al .
2004 , diemand et al .
2007 ) . in this context
, zentner et al . ( 2005 ) used semi - analytical models to claim that early formation times lead to significantly less substructure providing a natural explanation of the lack of galaxies in fossil groups .
these are the issues that we address in this paper .
we present n - body simulations of galaxy group halos to clarify the cosmic abundance and variance of substructure as related to magnitude gaps and formation history .
the plan of the paper is as follows . in
2 we present details of the numerical simulations and analysis procedure .
3 presents the results , whilst 4 summarizes our conclusions .
we analyze two @xmath21cdm simulations with cosmological parameters taken from the best - fits for wmap1 ( spergel et al .
2003 ) and wmap3 ( spergel et al .
these have present - day matter density parameters @xmath22=0.268 ( _ 0.238 _ for _ wmap3 _ ) ; cosmological constants contribution @xmath23=0.732(_0.762 _ ) ; baryonic contributions @xmath24=0.044 ( _ 0.042 _ ) ; and hubble parameters @xmath25(_0.73 _ ) ( @xmath26 km s@xmath27 mpc@xmath27 ) .
the mass fluctuation spectrum has a spectral index @xmath28(_0.951 _ ) , and is normalized by the linear rms fluctuation on @xmath29 , @xmath300.9(_0.75 _ ) .
the initial conditions are generated with grafic2 ( bertschinger 2001 ) .
all runs started at redshifts sufficiently high to ensure that the absolute maximum density contrast is still in the linear regime .
we followed the evolution of @xmath31 particles with masses of @xmath32 and @xmath33 for wmap1 and wmap3 in a box 90 mpc ( comoving ) on a side using the treecode pkdgrav ( stadel 2001 ) .
gravitational interactions between pairs of particles are softened in comoving coordinates with a spline softening length @xmath34 kpc ; forces are completely newtonian at twice this distance .
the particles have individual timesteps @xmath35 where @xmath36 is particle s acceleration .
the node - opening angle is @xmath37 after @xmath38 and @xmath39 earlier to provide higher force accuracy when the density is nearly uniform .
cell moments are expanded to fourth order . bound structures with a minimum of 250 particles at z=0
are identified using the spherical overdensity ( so ) algorithm ( macci et al .
2007 ) with a linking length @xmath40 equal to 0.2 times the mean interparticle separation . for each halo
, we fix the center at the most bound particle and find the mass within the virial radius where the spherical overdensity relative to the critical density for closure is @xmath41 and @xmath42 for wmap1 and wmap3 respectively ( mainini 2003 ) .
we discard structures with less than 250 particles and iterate the procedure to convergence .
if a particle is a potential member of two groups , it is assigned to the most massive one .
we select for our analysis all halos with masses in the range @xmath43 , resulting in 28 galaxy group sized halos for wmap1 and 16 groups for wmap3 with @xmath44 between 200,000 and 600,000 particles .
the subhalo population is identified with skid ( stadel 2001 ) , which calculates local densities using an sph kernel , then moves particles along the density gradient until they oscillate around a point characterized by some length @xmath40
. then particles are linked together using the friend - of - friend algorithm with a linking length equal to @xmath40 . using a linking length 4 times
the gravitational softening skid algorithm identifies the subhalo population inside each halo .
unfortunately by assuming this linking length the parent halo mass may be underestimated .
on the other hand using a larger linking length , 10 times the gravitational softening can cure this problem , but then some of the small sybila s are missed .
therefore we used a combination of the subhalo catalogs obtained with these two linking lengths in order to create the complete catalog of subhalos and to calculate their correct structural parameters .
we select for our analysis only subhalos with @xmath45 .
the assembly history of each halo may be studied using halo catalogs constructed at various times during the evolution of the system . for the purposes of our analysis , we concentrate on the period @xmath46 .
halos identified at @xmath47 are said to be _ progenitors _ of a @xmath48 system if at least @xmath49 of its particles are found within the virial radius of the latter . using this definition we can identify , at all times , the list of progenitors of a given @xmath48 galaxy group sized halo and track their properties through time . for each halo
, we use its mass assembly history ( mah ) to determine its formation time .
the mah is defined using the most massive progenitor and is usually well - fit by : @xmath50 , \label{eq : mah}\ ] ] with @xmath51 the redshift , @xmath52 the halo mass at present day and @xmath53 the formation redshift , defined as the redshift at which the fitted halo mah slope assumes the value of 2 ( wechsler et al . 2002 ) . only subhalos significantly larger than the gravitational softening survive the tidal forces of the parent halo . to test the scale
where this is significant , we re - simulated two galaxy group sized halos from the wmap1 simulation with 8 times higher mass resolution .
the test halos have 3,000,000 particles within the virial radius at z=0 .
figure [ fig : res ] shows that the low resolution ( solid lines ) and the high resolution runs ( dotted lines ) are in fairly good agreement for subhalos above our adopted resolution limit of @xmath54 .
we aim to select fossil groups in our sample of galaxy group sized halos .
ideally we want to determine a distribution of magnitude gaps and its relationship to the subhalo population in our sample .
since luminosity and magnitudes are not defined in our simulations , the identification of a magnitude gap between the first and second brightest member must be translated to a circular velocity gap .
if the faber - jackson ( 1976 ) relationship holds , then a 2 magnitude gap in the luminosity function corresponds to a circular velocity ratio @xmath55 , while a gap of 2.5 magnitudes would be @xmath56 . in observed systems , the brightest halo lies at the center of the x - ray emission as well as at the centroid of the redshift distribution , but its velocity dispersion is significantly less than the overall dispersion of the group , or equivalently , its x - ray temperature . the first brightest galaxy in our simulations is subsumed into the parent halo , so we need to calculate the ratio of @xmath57 that characterizes a fossil group .
in the case of rxj1340.6 + 4018 , its luminosity function has a magnitude gap of 2.5 , corresponding to the ratio of @xmath58 and @xmath59 , so that @xmath60 .
we adopt @xmath61 as a conservative criterion for a group with an appropriate magnitude gap .
we note that the local group and centaurus a group also meet this criterion and have substructure functions comparable to rxj1340.6 + 4018 .
subhalo mass determination in collisionless n - body simulations is uncertain . when dissipation is included in simulations , the subhalos number increases because the substructures become more robust and can survive to the very center of the parent halos ( nagai & kravtsov 2005 , macci et al .
hence we measured the circular velocities of substructure at the time of _ infall _ into the larger halos ( see also strigari et al .
after the subhalo falls into the parent halo it loses mass and its circular velocity decreases over time .
henceforth we compare the subhalo circular velocity at infall time with the observational data , not the subhalo circular velocities measured at present time .
figure [ fig : cos ] shows the circular velocity of the second most massive subhalo compared to that of the parent halo for our simulated groups .
the groups with @xmath62 are selected as candidate fossil groups .
only 18% of groups qualify , in fair agreement with the observed rate ( jones et al .
2003 , van den bosch et al .
but , does their substructure function look like that one of the fossil group rxj1340.6 + 4018 ?
we define a system to have a _ galaxy - like _ substructure function when it rescales to that of the milky way or equivalently rxj1340.6 + 4018 ( see e.g. figure 1 in donghia & lake 2004 ) . the opposite behavior would be a system with a _ cluster - like _ behavior , e.g. a rescaled virgo cluster .
the cumulative subhalo velocity distribution function of the halo sample is shown in figure [ fig : vdf ] .
note that in this figure we remove the most massive object ( i.e. the parent which should include the most massive galaxy ) from the cumulative substructure velocity distribution .
the top ( bottom ) panel is the simulations with wmap1 ( wmap3 ) cosmological parameters .
the candidate fossil groups are shown by dashed lines against a background of solid lines for the normal groups .
the panels show that the principal difference between the two realizations is that there are fewer group - sized halos with wmap3 parameters .
there is no statistically significant difference between the total number of subhalos with @xmath63 for candidate fossil groups and normal groups in the two realizations .
the cumulative substructure velocity distribution functions also show similar trends .
the range of cumulative substructure velocity distribution functions for all the galaxy group sized halos is shown in figure [ fig : desai ] as compared to the observed ones for galaxy clusters inferred from sdss data ( desai et al .
the agreement between predictions and observations is encouraging , although the observed cumulative distribution exhibits a broader range for @xmath64 from 0.25 to 0.8 , whereas the second most massive subhalo in the simulations is never larger than 50% of the parent halo .
the larger spread might owe to uncertainties in the estimate of the parent s halo velocity dispersion from a relatively small number of member velocities .
the trend of the cumulative subhalo velocity function in simulations agrees well with the one inferred from galaxy clusters .
note that figure [ fig : desai ] shows that there are galaxy clusters with the second most massive member having @xmath65 25 - 30% of the parent halo , showing evidence for magnitude gaps in sdss data .
figure [ fig : obs ] shows the range of cumulative subhalo velocity distribution functions for candidate fossil groups in simulations , marked by shaded areas , as compared to the observed ones for rxj1340.6 + 4018 , the local group and centaurus a ( donghia & lake 2004 for a compilation and reference therein ) .
the local group uses data from mateo et al .
( 1998 ) , odenkirchen et al .
( 2001 ) and kleyna et al .
( 2005 ) ; we used the results of kazantzidis et al .
( 2004 ) to convert three - dimensional velocity dispersion into circular velocity .
we considered the number of satellites per central galaxy instead of considering the local group as a whole .
this makes the comparison with simulations straightforward ( see macci et al .
2006b for more details ) . note that while we find systems with appropriate gaps in velocity between the second most massive subhalo and the most massive one , the cosmic variance in cdm simulations is insufficient to explain the cumulative number of subhalos with circular velocity less than 20% of observed systems ( e.g. @xmath66 ) .
numerical simulations performed with wmap1 and wmap3 parameters give almost identical results ; the narrower region covered by the velocity distribution function in wmap3 owes to the lower number of candidate fossil groups .
these systems populate the lower region of the shaded area shown in figure [ fig : desai ] .
the slope of the function agrees well with the one inferred from normal groups and galaxy clusters . yet
, both wmap1 and wmap3 models over - predict the total number of substructures relative to centaurus a and rxj1340.6 + 401 . especially for this latter one the discrepancy in the velocity range of subhalos with circular velocity larger than 10%
is a factor of six .
can this discrepancy owe to uncertainties in the estimate of the total mass ( circular velocity ) of the parent halo of rxj1340.6 + 401 ? if the total mass is lower than estimated km / s with great uncertainties due to the low number of galaxy members ( jones et al .
the points in figure [ fig : obs ] may shift on the right leading to a fair agreement with the expectations ( dashed area ) .
however the x - ray temperature is @xmath0 1 kev for the archetype fossil group , meaning a circular velocity of the parent halo at least 580 km / s , so the points ca nt be shifted to the right to obtain a better agreement .
also , dynamical friction and merging are not a general solution to the discrepancy .
these dynamical effects are included in the full numerical simulations .
any specific substructure function evolves in the same way by dynamical friction and merging independent of the parent mass ( donghia & lake 2004 ) .
dynamical friction can be important in promoting the merger of the largest objects in less than one hubble time but dynamical friction alone will not create substructure functions that are different for different parent masses . clearly galaxies could have had a long time to evolve by dynamical friction , but this is also included in the simulations and does not affect the substructure function below @xmath12 of 0.2 .
figure [ fig : nv2vp ] shows that the total number of substructures with ( @xmath67 ) in our simulations is independent of any velocity gap : @xmath64 . yet , we do nt find a halo population as poor in substructure as rxj1340.6 + 4018 and the local group .
recently , using semi - analytic models , zentner et al .
( 2005 ) have highlighted the possibility that early forming halos may have significant less substructures , providing a potential explanation of the lack of substructures in the observed fossil groups .
we explore this claim in figure [ fig : n01 ] , where the total number of subhalos with circular velocity larger than 10% of the parent halo is shown as a function of the formation redshift ( @xmath53 ) of the parent halo and the two sets of cosmological parameters .
both panels show that there is an anticorrelation between the halo formation redshift and the total number of substructures .
halos that form earlier exhibit a lower number of satellites within a factor two for a cosmology with wmap1 parameters and slightly larger and with more spread adopting wmap3 parameters .
we tested also the distribution of the halo formation redshift of all the halos selected in the wmap1 and wmap3 simulations .
fossil groups and normal groups in the range of mass of @xmath68m@xmath19 here adopted tend to form in a narrow range of redshift centered around z=1.5 and z=0.9 for cosmologies with wmap1 and wmap3 parameters , respectively .
the candidate fossil group halos tend to form slightly earlier than the normal groups ones , even if the sample of fossil groups is limited .
is this the solution to the missing galaxy question in rxj1340.6 + 4018 ? in figure [ fig : n01 ] we draw three additional horizontal lines tracing the total number of observed galaxies with circular velocity larger than 10% of the parent halo velocity ( @xmath69 ) for the galaxy group centaurus a , rxj1340.6 + 4018 and the local group .
while the early formation redshift can reduce on average of a factor two the total number of satellites in simulations , the observations show a discrepancy greater than a factor of five .
_ we conclude that the early formation time per se can reduce but not reconcile the discrepancy between the observed total number of satellites to the number predicted by numerical simulations in galaxy groups .
_ is the anticorrelation between total number of satellites and redshift formation exhibited in figure 6 in conflict with the previous finding of self - similarity of the subhalo population of dark matter halos ? in order to assess this effect we analyzed separately for each halo the _ normalization _ of the cumulative velocity distribution function and its slope as a function of the formation redshift .
the substructure function of a all halos is known to be approximated by the following power law : @xmath70 with @xmath71 over a wide range of mass and cosmological models ( klypin 1999 , klypin 2003 ) . for each halo
we fit the power law in the range of @xmath72 between 0.1 and 0.3 in order to to avoid to bias the fossil group candidates in respect to the normal group halos , since the first ones have , by definition a lack of subhalos for @xmath73 . during the fitting procedure the normalization @xmath74 and
the slope @xmath75 are treated as free parameters .
the values and associated uncertainties of @xmath74 and @xmath75 are determined via a @xmath76 minimization procedure using the lavenberg & marquart method .
the slope distribution of all the halos selected at z=0 is shown by the histogram in the left panel of figure [ fig : slopes ] .
the panel on the right of the figure [ fig : slopes ] shows the lack of correlation between the slope @xmath75 of the substructure function of each halo and its formation redshift @xmath53 .
_ we conclude that only the normalization of the cumulative velocity distribution function of satellites anticorrelates with the formation time , the slope does not_. this result highlights that non - linear structures in cdm models are self - similar , independent of the formation time and insensitive to the difference in cosmological parameters used in the two simulations .
thus , while the slope of the cumulative substructure function is set by the cosmological model ( setting the self similarity of dark halos ) , its normalization depends on the local environment and on the formation history of the halo .
we draw additional horizontal lines in figure [ fig : slopes ] showing the slope inferred for the local group , rxj1340.6 + 4018 and centaurus a and sdss galaxy clusters ( the middle two are so close that they are hard to resolve ) .
the sdss galaxy cluster slope quoted by desai et al .
( 2004 ) is @xmath77 in fair agreement with numerical simulations .
this shallower value quoted for sdss data is because it is estimated over the entire range of @xmath12 , whereas the slope of the simulations is obtained in a narrower range ( for [email protected] ) .
however , the observed slopes of galaxy groups like centaurus a and rxj1340.6 + 4018 are substantially lower .
rxj1340.6 + 4018 has a typical mass of 25% of the virgo cluster mass , but it exibits a substructure population similar to the rescaled version of the local group .
we used numerical simulations of group sized halos to examine the formation of systems with paucity of substructures like the fossil system rxj1340.6 + 4018 .
our results show that : * in the mass range 1 - 5@xmath79m@xmath19 almost 18% of the group sized halos are candidate fossil groups according to the definition of velocity gap , but not a single system has a shallow cumulative substructure function similar to that one exhibited from the local group or rxj1340.6 + 4018 .
* current observations of galaxy groups with a velocity gap exhibit a paucity of substructures with a substructure function like that of the local group , whereas the total number of substructures predicted from cdm models is almost constant and independent of any gap in velocity . *
the total number of substructures identified at the present time does indeed correlate with the formation time of the halo , as found in semi - analytics work .
however , the early formation time reduces the average number of substructures by a factor two , while a factor of six is needed to reconcile the predictions with the observations .
we showed that only the normalization of the cumulative velocity function of satellites depends on the formation time , whereas the slope does not .
this seems intuitive as the fluctuations in cdm models are gaussian with random phases . while the early formation of a parent halo from a large scale wave will bias the formation of substructures ,
it should not and indeed does not tilt the smaller scale spectrum .
* we find that the cumulative velocity distribution function of substructures is self - similar in cdm models , with the slope being the same in the wmap1 and wmap3 realizations and they agree with previous results on halos with galaxy masses and galaxy clusters .
yet , only the slope of the cumulative velocity function of substructures in the sdss galaxy cluster sample agrees with the predictions of the numerical simulations , while the slopes of groups of galaxies like the local group , cenaturus a or the rxj1340.6 + 4018 are significantly shallower .
we explored the effects of cosmic variance on the cumulative substructure function and find that the deviation of the model predictions from the observations begins for subhalos with circular velocity larger than a 20% of the parent halo circular velocity .
for a typical galaxy group sized halo with circular velocity of @xmath80 , a missing subhalo in the observations has a circular velocity of @xmath81 . physical mechanisms invoking reionization or feedback processes from satellites are already stretching to suppress halos at @xmath13 . here
we showed that invoking a earlier formation time for these objects may alleviate the problem but can not reproduce the observed number of visible galaxies in groups of galaxies with intermidiate mass between galaxies and poor cluster of galaxies .
if future spectroscopic luminosity function for x - ray emitting groups in the range of mass intermidiate between the local group and the virgo cluster produces a larger sample of groups with paucity of substructures this class of objects will remain an interesting challenge for the current model of structure formation .
the numerical simulations were performed on zbox2 supercomputer at the university of zurich ; ( http://www-theorie.physik.unizh.ch/@xmath0dpotter/zbox2/ ) .
thanks highlighting discussions with daniele pierini , maurilio pannella and stefano zibetti on observational data . e.d .
is financed by a eu marie curie intra - european fellowship under contract meif-041569 . | we use large cosmological n - body simulations to study the subhalo population in galaxy group sized halos . in particular , we look for fossil group candidates with typical masses @xmath0 10 - 25% of virgo cluster but with an order of magnitude less substructure .
we examine recent claims that the earliest systems to form are deficient enough in substructure to explain the luminosity function found in fossil groups .
although our simulations show a correlation between the halo formation time and the number of subhalos , the maximum suppression of subhalos is a factor of 2 - 2.5 , whereas a factor of 6 is required to match fossil groups and galaxies .
while the number of subhalos depends weakly on the formation time , the slope of the halo substructure velocity function does not . the satellite population within cold dark matter ( cdm ) halos is self - similar at scales between galaxies and galaxy clusters regardless of mass , whereas current observations show a break in self - similarity at a mass scale corresponding to group of galaxies .
psfig.sty = 10000 # 1to 0pt#1 # 1#2#1 8 _ 8 @xmath1 2 r_-2 @xmath2 2 c_-2 @xmath3 200 r_200 @xmath4 200 v_200 @xmath5 200 m_200 @xmath6
0 _ 0 @xmath7 200@xmath8 [ firstpage ] galaxies : formation galaxies : halos galaxies : structure cosmology : theory dark matter large - scale structure of universe methods : numerical , n - body simulation |
* ( a ) * a countable borel equivalence relation on a standard borel space @xmath2 is a borel equivalence relation @xmath3 with the property that every equivalence class @xmath4_e$ ] , @xmath5 , is countable .
we denote by @xmath6 the class of countable borel equivalence relations . over the last 25 years
there has been an extensive study of countable borel equivalence relations and their connection with group actions and ergodic theory .
an important aspect of this work is an understanding of the kind of countable ( first - order ) structures that can be assigned in a uniform borel way to each class of a given equivalence relation .
this is made precise in the following definitions ; see @xcite , section 2.5 .
let @xmath7 be a countable relational language , where @xmath8 has arity @xmath9 , and @xmath0 a class of countable structures in @xmath10 closed under isomorphism .
let @xmath1 be a countable borel equivalence relation on a standard borel space @xmath2 .
an @xmath10-structure on @xmath1 is a borel structure @xmath11 of @xmath10 with universe @xmath2 ( i.e. , each @xmath12 is borel ) such that for @xmath13 and @xmath14 , @xmath15 .
then each @xmath1-class @xmath16 is the universe of the countable @xmath10-structure @xmath17 .
if for all such @xmath18 , we say that @xmath19 is a @xmath0-structure on @xmath1 .
finally if @xmath1 admits a @xmath0-structure , we say that @xmath1 is @xmath0-structurable .
many important classes of countable borel equivalence relations can be described as the @xmath0-structurable relations for appropriate @xmath0 .
for example , the hyperfinite equivalence relations are exactly the @xmath0-structurable relations , where @xmath0 is the class of linear orderings embeddable in @xmath20 .
the treeable equivalence relations are the @xmath0-structurable relations , where @xmath0 is the class of countable trees ( connected acyclic graphs ) .
the equivalence relations generated by a free borel action of a countable group @xmath21 are the @xmath0-structurable relations , where @xmath0 is the class of structures corresponding to free transitive @xmath21-actions .
the equivalence relations admitting no invariant probability borel measure are the @xmath0-structurable relations , where @xmath22 , @xmath23 unary and @xmath24 binary , and @xmath0 consists of all countably infinite structures @xmath25 , with @xmath26 an infinite , co - infinite subset of @xmath27 and @xmath28 the graph of a bijection between @xmath27 and @xmath26 . for @xmath7 as before and countable set @xmath2 , we denote by @xmath29 the standard borel space of countable @xmath10-structures with universe
. clearly every countable @xmath10-structure is isomorphic to some @xmath30 , for @xmath31 . given a class @xmath0 of countable @xmath10-structures , closed under isomorphism , we say that @xmath0 is borel if @xmath32 is borel in @xmath29 , for each countable set @xmath2 .
we are interested in borel classes @xmath0 in this paper . for any @xmath33-sentence @xmath34 , the class of countable models of @xmath34 is borel . by a classical theorem of lopez - escobar @xcite ,
every borel class @xmath0 of @xmath10-structures is of this form , for some @xmath35-sentence @xmath34 .
we will also refer to such @xmath34 as a theory . adopting this model - theoretic point of view , given a theory @xmath34 and a countable borel equivalence relation @xmath1
, we put @xmath36 if @xmath1 is @xmath0-structurable , where @xmath0 is the class of countable models of @xmath34 , and we say that @xmath1 is @xmath34-structurable if @xmath37 .
we denote by @xmath38 the class of @xmath34-structurable countable borel equivalence relations .
finally we say that a class @xmath39 of countable borel equivalence relations is elementary if it is of the form @xmath40 , for some @xmath34 ( which axiomatizes @xmath39 ) . in some sense
the main goal of this paper is to study the global structure of elementary classes .
first we characterize which classes of countable borel equivalence relations are elementary .
we need to review some standard concepts from the theory of borel equivalence relations . given equivalence relations @xmath41 on standard borel spaces @xmath42 , resp .
, a borel homomorphism of @xmath1 to @xmath43 is a borel map @xmath44 with @xmath45 .
we denote this by @xmath46 .
if moreover @xmath47 is such that all restrictions @xmath48_e \colon [ x]_e \to [ f(x)]_f$ ] are bijective , we say that @xmath47 is a class - bijective homomorphism , in symbols @xmath49 .
if such @xmath47 exists we also write @xmath50 .
we similarly define the notion of class - injective homomorphism , in symbols @xmath51 .
a borel reduction of @xmath1 to @xmath43 is a borel map @xmath44 with @xmath52 .
we denote this by @xmath53 .
if @xmath47 is also injective , it is called a borel embedding , in symbols @xmath54 . if there is a borel reduction of @xmath1 to @xmath43 we write @xmath55 and if there is a borel embedding we write @xmath56 .
an invariant borel embedding is a borel embedding @xmath47 as above with @xmath57 @xmath43-invariant .
we use the notation @xmath58 and @xmath59 for these notions . by the usual schroeder - bernstein argument , @xmath60 , where @xmath61 is borel isomorphism .
kechris - solecki - todorcevic @xcite proved a universality result for theories of graphs , which was then extended to arbitrary theories by miller ; see corollary [ thm : einftysigma ] .
[ thm:11 ] for every theory @xmath34 , there is an invariantly universal @xmath34-structurable countable borel equivalence relation @xmath62 , i.e. , @xmath63 , and @xmath64 for any other @xmath65 .
clearly @xmath62 is uniquely determined up to borel isomorphism .
in fact in theorem [ thm : esigma ] we formulate a relative " version of this result and its proof that allows us to capture more information .
next we note that clearly every elementary class is closed downwards under class - bijective borel homomorphisms .
we now have the following characterization of elementary classes ( see corollary [ thm : elem - char ] ) .
[ thm:12 ] a class @xmath66 of countable borel equivalence relations is elementary iff it is ( downwards-)closed under class - bijective homomorphisms and contains an invariantly universal element @xmath67 .
examples of non - elementary classes include the class of non - smooth countable borel equivalence relations ( a countable borel equivalence relation is smooth if it admits a borel transversal ) , the class of equivalence relations admitting an invariant borel probability measure , and the class of equivalence relations generated by a free action of _ some _ countable group .
more generally , nontrivial unions of elementary classes are never elementary ( see [ thm : union - nonelem ] ) .
next we show that every @xmath68 is contained in a ( unique ) smallest ( under inclusion ) elementary class ( see [ thm : elem - cone ] ) .
[ thm:13 ] for every @xmath69 , there is a smallest elementary class containing @xmath1 , namely @xmath70 . many classes of countable borel equivalence relations that have been extensively studied , like hyperfinite or treeable ones , are closed ( downwards ) under borel reduction .
it turns out that every elementary class is contained in a ( unique ) smallest ( under inclusion ) elementary class closed under borel reduction ( see [ thm : red - elem ] ) .
[ thm:14 ] for every elementary class @xmath39 , there is a smallest elementary class containing @xmath39 and closed under borel reducibility , namely @xmath71 .
we call an elementary class closed under reduction an elementary reducibility class . in analogy with [ thm:12 ] , we have the following characterization of elementary reducibility classes ( see [ thm : elemred - char ] ) .
below by a smooth borel homomorphism of @xmath72 into @xmath73 we mean a borel homomorphism for which the preimage of any point is smooth for @xmath1 .
[ thm:15 ] a class @xmath66 is an elementary reducibility class iff it is closed ( downward ) under smooth homomorphisms and contains an invariantly universal element @xmath67 .
we note that as a corollary of the proof of [ thm:14 ] every elementary reducibility class is also closed downward under class - injective borel homomorphisms .
hjorth - kechris ( * ? ? ?
* d.3 ) proved ( in our terminology and notation ) that every @xmath74 ( @xmath39 elementary ) is closed under @xmath75 , i.e. , containment of equivalence relations on the same space . since containment is a class - injective homomorphism ( namely the identity ) , [ thm:14 ] generalizes this .
we also prove analogous results for borel embeddability instead of borel reducibility ( see [ thm : emb - elem ] ) .
for each countably infinite group @xmath21 denote by @xmath76 the elementary class of equivalence relations induced by free borel actions of @xmath21 .
its invariantly universal element is the equivalence relation induced by the free part of the shift action of @xmath21 on @xmath77 .
for trivial reasons this is not closed under borel reducibility , so let @xmath78 be the elementary class of all equivalence relations whose aperiodic part is in @xmath76
. then we have the following characterization ( see [ thm : freeact ] ) . [ thm:16 ]
let @xmath21 be a countably infinite group
. then the following are equivalent : 1 .
@xmath78 is an elementary reducibility class .
@xmath21 is amenable .
we call equivalence relations of the form @xmath79 universally structurable . denote by @xmath80 the class of universally structurable equivalence relations . in view of [ thm:11 ] , an elementary class
is uniquely determined by its invariantly universal such equivalence relation , and the poset of elementary classes under inclusion is isomorphic to the poset @xmath81 of borel isomorphism classes of universally structurable equivalence relations under invariant borel embeddability .
it turns out that this poset has desirable algebraic properties ( see [ thm : lattice ] ) .
[ thm:17 ] the poset @xmath82 is an @xmath83-complete , distributive lattice .
moreover , the inclusion @xmath84 preserves ( countable ) meets and joins .
this has implications concerning the structure of the class of universally structurable equivalence relations under borel reducibility .
the order - theoretic structure of the poset @xmath85 of _ all _ bireducibility classes under @xmath86 is not well - understood , apart from that it is very complicated ( by @xcite ) .
the first general study of this structure was made only recently by kechris - macdonald in @xcite .
in particular , they pointed out that it was even unknown whether there exists _ any _ pair of @xmath86-incomparable @xmath87 for which a @xmath86-meet exists .
however it turns out that the subposet @xmath88 behaves quite well ( see [ thm : lattice - red ] ) .
[ thm:18 ] the poset of universally structurable bireducibility classes , under @xmath86 , @xmath88 is an @xmath83-complete , distributive lattice .
moreover , the inclusion into the poset @xmath85 of all bireducibility classes , under @xmath86 , preserves ( countable ) meets and joins . adapting the method of adams - kechris @xcite
, we also show that this poset is quite rich ( see [ thm : pbr ] ) .
[ thm:19 ] there is an order - embedding from the poset of borel subsets of @xmath89 under inclusion into the poset @xmath88 .
the combination of [ thm:18 ] and [ thm:19 ] answers the question mentioned in the paragraph following [ thm:17 ] by providing a large class of @xmath90-incomparable countable borel equivalence relations for which @xmath86-meets exist .
an important question concerning structurability is which properties of a theory @xmath91 yield properties of the corresponding elementary class @xmath40 .
the next theorem provides the first instance of such a result .
marks ( * ? ? ?
* end of section 4.3 ) asked ( in our terminology ) for a characterization of when the elementary class @xmath92 , where @xmath93 is a scott sentence of a countable structure , consists of smooth equivalence relations , or equivalently , when @xmath94 is smooth .
we answer this problem in full generality , i.e. , for an arbitrary theory @xmath34 .
although this result belongs purely in the category of borel equivalence relations , our proof uses ideas and results from topological dynamics and ergodic theory ( see [ thm : einftys - smooth ] ) .
[ thm:110 ] let @xmath34 be a theory .
the following are equivalent : 1 .
@xmath40 contains only smooth equivalence relations , i.e. , @xmath62 is smooth .
2 . there is an @xmath33-formula @xmath95 which defines a finite nonempty subset in any countable model of @xmath34 .
along these lines an interesting question is to find out what theories @xmath34 have the property that every aperiodic countable borel equivalence relation is @xmath34-structurable .
a result that some particular @xmath34 axiomatizes all aperiodic @xmath1 shows that every such @xmath69 carries a certain type of structure , which can be useful in applications .
a typical example is the very useful marker lemma ( see @xcite ) , which shows that every aperiodic @xmath1 admits a decreasing sequence of borel complete sections @xmath96 with empty intersection .
this can be phrased as : every aperiodic countable borel equivalence relation @xmath1 is @xmath34-structurable , where @xmath34 in the language @xmath97 asserts that each ( unary ) @xmath98 defines a nonempty subset , @xmath99 , and @xmath100 .
a particular case is when @xmath101 is a scott sentence of a countable structure . for convenience
we say that @xmath1 is @xmath19-structurable if @xmath1 is @xmath93-structurable .
marks recently pointed out that the work of @xcite implies a very general condition under which this happens ( see [ thm : einftya - univ ] ) .
[ thm:111 ] let @xmath19 be a countable structure with trivial definable closure .
then every aperiodic countable borel equivalence relation is @xmath19-structurable .
in particular ( see [ thm:817 ] ) the following frass structures can structure every aperiodic countable borel equivalence relation : @xmath102 , the random graph , the random @xmath103-free graph ( where @xmath103 is the complete graph on @xmath104 vertices ) , the random poset , and the rational urysohn space .
finally we mention an application to a model - theoretic question that has nothing to do with equivalence relations
. the concept of amenability of a structure in the next result ( see [ thm : amenable - ntdc ] ) can be either the one in @xcite ) or the one in @xcite .
this result was earlier proved by the authors by a different method ( still using results of @xcite ) but it can also be seen as a corollary of [ thm:111 ] .
let @xmath19 be a countably infinite amenable structure .
then @xmath19 has non - trivial definable closure .
* ( b ) * this paper is organized as follows : in [ sec : prelims ] we review some basic background in the theory of borel equivalence relations and model theory . in [ sec : struct - equiv ] we introduce the concept of structurability of equivalence relations and discuss various examples . in [ sec : univcons ] we study the relationship between structurability and class - bijective homomorphisms , obtaining the tight correspondence given by ; we also apply structurability to describe a product construction ( class - bijective or `` tensor '' product ) between countable borel equivalence relations . in [ sec : reductions ] we study the relationship between structurability and other kinds of homomorphisms , such as reductions ; we also consider the relationship between reductions and compressible equivalence relations . in [ sec : poset ] we introduce some concepts from order theory convenient for describing the various posets of equivalence relations we are considering , and then study the poset @xmath82 of universally structurable equivalence relations ( equivalently of elementary classes ) . in [ sec : freeact ] we consider the elementary class @xmath76 of equivalence relations induced by free actions of a countable group @xmath21 . in
[ sec : struct - logic ] we consider relationships between model - theoretic properties of a theory @xmath34 and the corresponding elementary class @xmath40 . finally , in [ sec : open - problems ] we list several open problems related to structurability . the appendices contain some generalizations , alternative points of view , and miscellaneous results which are tangential to the main subject of this paper . in [ sec : fiber ] we introduce fiber spaces ( previously considered in @xcite and @xcite ) , discuss their category - theoretic aspects , and discuss generalizations to that context of several concepts appearing in the body of this paper ( including structurability and the various kinds of homomorphisms ) . in [ sec : interp ] we introduce a categorical structure on the class of all theories which interacts well with structurability . finally , in [ sec : sdlat ] we prove a lattice - theoretic result generalizing the well - known loomis - sikorski representation theorem for @xmath34-boolean algebras , which can be applied in particular to the lattice @xmath82 considered in [ sec : lattice ] .
we would like to thank andrew marks for many valuable suggestions and for allowing us to include [ thm:111 ] in this paper .
for general model theory , see @xcite . for general classical descriptive set theory , see @xcite . by a language , we will always mean a countable first - order relational language , i.e. , a countable set @xmath105 of relation symbols , where each @xmath8 has an associated arity @xmath106 . the only logic we will consider is the infinitary logic @xmath33 .
we use letters like @xmath107 for formulas , and @xmath108 for sentences . by a theory , we mean a pair @xmath109 where @xmath10 is a language and @xmath34 is an @xmath33-sentence .
when @xmath10 is clear from context , we will often write @xmath34 instead of @xmath109 . let @xmath10 be a language . by an @xmath10-structure ,
we mean in the usual sense of first - order logic , i.e. , a tuple @xmath110 where @xmath2 is a set and for each @xmath104-ary relation symbol @xmath111 , @xmath112 is an @xmath104-ary relation on @xmath2 .
then as usual , for each formula @xmath113 with @xmath104 free variables , we have an interpretation @xmath114 as an @xmath104-ary relation on @xmath2
. we write @xmath29 for the set of @xmath10-structures with universe @xmath2 . more generally , for a theory @xmath109 , we write @xmath115 for the set of models of @xmath34 with universe @xmath2 . when @xmath2 is countable , we equip @xmath115 with its usual standard borel structure ( see e.g. , @xcite ) .
if @xmath110 is an @xmath10-structure and @xmath116 is a bijection , then we write @xmath117 for the pushforward structure , with universe @xmath118 and @xmath119 for @xmath104-ary @xmath23 and @xmath120 .
when @xmath121 , this defines the logic action of @xmath122 ( the group of bijections of @xmath2 ) on @xmath123 .
if @xmath124 is any function , then @xmath125 is the pullback structure , with universe @xmath118 and @xmath126 when @xmath47 is the inclusion of a subset @xmath127 , we also write @xmath128 for @xmath125 .
every countable @xmath10-structure @xmath19 has a scott sentence @xmath93 , which is an @xmath33-sentence whose countable models are exactly the isomorphic copies of @xmath19 .
a borel class of @xmath10-structures is a class @xmath0 of _ countable _ @xmath10-structures which is closed under isomorphism and such that @xmath129 is borel for every countable set @xmath2 ( equivalently , for @xmath130 ) .
for example , for any @xmath33-sentence @xmath34 , the class of countable models of @xmath34 is borel . by a classical theorem of lopez - escobar @xcite ,
every borel class @xmath0 of @xmath10-structures is of this form , for some @xmath34 .
( while lopez - escobar s theorem is usually stated only for @xmath131 , it is easily seen to hold also for @xmath29 with @xmath2 finite . )
a borel equivalence relation @xmath1 on a standard borel space @xmath2 is an equivalence relation which is borel as a subset of @xmath132 ; the equivalence relation @xmath1 is countable if each of its classes is .
we will also refer to the pair @xmath133 as an equivalence relation . throughout this paper
, we use @xmath6 to denote the class of countable borel equivalence relations @xmath133 .
if @xmath21 is a group acting on a set @xmath2 , then we let @xmath134 be the orbit equivalence relation : @xmath135 if @xmath21 is countable , @xmath2 is standard borel , and the action is borel , then @xmath136 is a countable borel equivalence relation .
conversely , by the feldman - moore theorem @xcite , every countable borel equivalence relation on a standard borel space @xmath2 is @xmath136 for some countable group @xmath21 with some borel action on @xmath2 .
if @xmath21 is a group and @xmath2 is a set , the ( right ) shift action of @xmath21 on @xmath137 is given by @xmath138 for @xmath139 , @xmath140 , and @xmath141 .
we let @xmath142 denote the orbit equivalence of the shift action .
if @xmath21 is countable and @xmath2 is standard borel , then @xmath143 is a countable borel equivalence relation .
if @xmath21 already acts on @xmath2 , then that action embeds into the shift action , via @xmath144 in particular , any action of @xmath21 on a standard borel space embeds into the shift action of @xmath21 on @xmath77 .
the free part of a group action of @xmath21 on @xmath2 is @xmath145 the action is free if the free part is all of @xmath2 .
we let @xmath146 denote the orbit equivalence of the free part of the shift action of @xmath21 on @xmath137 .
an invariant measure for a borel group action of @xmath21 on @xmath2 is a nonzero @xmath34-finite borel measure @xmath147 on @xmath2 such that @xmath148 for all @xmath139 ( where @xmath149 is the pushforward ) .
an invariant measure on a countable borel equivalence relation @xmath133 is an invariant measure for some borel action of a countable group @xmath21 on @xmath2 which generates @xmath1 , or equivalently for any such action ( see @xcite ) .
an invariant measure @xmath147 on @xmath133 is ergodic if for any @xmath1-invariant borel set @xmath150 , either @xmath151 or @xmath152 .
let @xmath153 be countable borel equivalence relations , and let @xmath116 be a borel map ( we write @xmath154 to denote that @xmath47 is borel ) .
we say that @xmath47 is : * a homomorphism , written @xmath155 , if @xmath156 i.e. , @xmath47 induces a map on the quotient spaces @xmath157 ; * a reduction , written @xmath158 , if @xmath47 is a homomorphism and @xmath159 i.e. , @xmath47 induces an injection on the quotient spaces ; * a class - injective homomorphism ( respectively , class - surjective , class - bijective ) , written @xmath160 ( respectively @xmath161 , @xmath162 ) , if @xmath47 is a homomorphism such that for each @xmath163 , the restriction @xmath48_e : [ x]_e - > [ f(x)]_f$ ] to the equivalence class of @xmath164 is injective ( respectively , surjective , bijective ) ; * an embedding , written @xmath165 , if @xmath47 is an injective ( or equivalently , class - injective ) reduction ; * an invariant embedding , written @xmath166 , if @xmath47 is a class - bijective reduction , or equivalently an embedding such that the image @xmath167 is @xmath43-invariant . among these various kinds of homomorphisms ,
the reductions have received the most attention in the literature , while the class - bijective ones are most closely related to the notion of structurability . here is a picture of the containments between these classes of homomorphisms , with the more restrictive classes at the bottom : @xmath168 \node(invemb ) { \sqle_b^i } ; \node(emb ) at ( [ shift={(-1,1)}]invemb ) { \sqle_b } edge ( invemb ) ; \node(cb ) at ( [ shift={(1,1)}]invemb ) { ->_b^{cb } } edge ( invemb ) ; \node(red ) at ( [ shift={(-1,1)}]emb ) { \le_b } edge ( emb ) ; \node(ci ) at ( [ shift={(0,2)}]invemb ) { ->_b^{ci } } edge ( emb ) edge ( cb ) ; \node(cs ) at ( [ shift={(1,1)}]cb ) { ->_b^{cs } } edge ( cb ) ; \node(hom ) at ( [ shift={(0,1.5)}]ci ) { ->_b } edge ( red ) edge ( ci ) edge ( cs ) ; \end{tikzpicture}\ ] ] we say that @xmath133 ( borel ) reduces to @xmath169 , written @xmath170 ( or simply @xmath171 ) , if there is a borel reduction @xmath158 .
similarly for the other kinds of homomorphisms , e.g. , @xmath1 embeds into @xmath43 , written @xmath172 , if there is some @xmath173 , etc .
we also write : * @xmath174 ( @xmath1 is bireducible to @xmath43 ) if @xmath171 and @xmath175 ; * @xmath176 if @xmath171 and @xmath177 , and similarly for @xmath178 and @xmath179 ; * @xmath180 ( @xmath1 is class - bijectively equivalent to @xmath43 ) if @xmath181 and @xmath182 ; * @xmath183 if @xmath1 is borel isomorphic to @xmath43 , or equivalently ( by the borel schrder - bernstein theorem ) @xmath184 and @xmath185 . clearly @xmath86 , @xmath186 , @xmath187 , etc . ,
are preorders on the class @xmath6 , and @xmath188 , @xmath189 , @xmath61 are equivalence relations on @xmath6 .
the @xmath188-equivalence classes are called bireducibility classes , etc . we have the following basic operations on borel equivalence relations .
let @xmath190 be borel equivalence relations .
their disjoint sum is @xmath191 where @xmath192 is the disjoint union of @xmath193 , and @xmath194 relates elements of @xmath2 according to @xmath1 and elements of @xmath118 according to @xmath43 and does not relate elements of @xmath2 with elements of @xmath118 .
the canonical injections @xmath195 and @xmath196 are then invariant embeddings @xmath197 .
clearly the disjoint sum of countable equivalence relations is countable .
we have obvious generalizations to disjoint sums of any countable family of equivalence relations .
their cross product is @xmath198 , where @xmath199 ( the `` cross '' adjective is to disambiguate from the _ tensor products _ to be introduced in [ sec : tensor ] . )
the projections @xmath200 and @xmath201 are class - surjective homomorphisms @xmath202 .
cross products also generalize to countably many factors ; but note that only _
finite _ cross products of countable equivalence relations are countable .
recall that an equivalence relation @xmath133 is countable if each @xmath1-class is countable ; similarly , it is finite if each @xmath1-class is finite , and aperiodic countable if each @xmath1-class is countably infinite .
a countable borel equivalence relation is always the disjoint sum of a finite borel equivalence relation and an aperiodic countable borel equivalence relation .
since many of our results become trivial when all classes are finite , we will often assume that our equivalence relations are aperiodic .
for any set @xmath2 , the indiscrete equivalence relation on @xmath2 is @xmath203 .
a borel equivalence relation @xmath133 is smooth if @xmath204 where @xmath205 is the equality relation on some standard borel space @xmath118 .
when @xmath1 is countable , this is equivalent to @xmath1 having a borel transversal , i.e. , a borel set @xmath150 meeting each @xmath1-class exactly once , or a borel selector , i.e. , a borel map @xmath206 such that @xmath207 and @xmath208 for all @xmath209 .
any finite borel equivalence relation is smooth .
up to bireducibility , the smooth borel equivalence relations consist exactly of @xmath210 and these form an initial segment of the preorder @xmath211 ( silver s dichotomy ; see @xcite ) .
we will sometimes use the standard fact that a countable borel equivalence relation @xmath133 is smooth iff every ergodic invariant ( @xmath34-finite borel ) measure on @xmath1 is atomic .
( for the converse direction , use e.g. , that if @xmath1 is not smooth , then @xmath212 ( see [ thm : hyperfinite ] below ) , and @xmath213 is isomorphic to the orbit equivalence of the translation action of @xmath214 on @xmath89 , which admits lebesgue measure as an ergodic invariant nonatomic @xmath34-finite measure . ) if @xmath116 is any function between sets , the kernel of @xmath47 is the equivalence relation @xmath215 on @xmath2 given by @xmath216 .
so a borel equivalence relation is smooth iff it is the kernel of some borel map .
a countable borel equivalence relation @xmath1 is universal if @xmath1 is @xmath86-greatest in @xmath6 , i.e. , for any other countable borel equivalence relation @xmath43 , we have @xmath175 .
an example is @xmath217 ( where @xmath218 is the free group on @xmath219 generators ) @xcite .
a countable borel equivalence relation @xmath1 is invariantly universal if @xmath1 is @xmath220-greatest in @xmath6 , i.e. , for any other countable borel equivalence relation @xmath43 , we have @xmath185 .
we denote by @xmath221 any such @xmath1 ; in light of the borel schrder - bernstein theorem , @xmath221 is unique up to isomorphism .
clearly @xmath221 is also @xmath86-universal .
( note : in the literature , @xmath221 is commonly used to denote any @xmath86-universal countable borel equivalence relation ( which is determined only up to bireducibility ) . )
one realization of @xmath221 is @xmath222 .
( this follows from the feldman - moore theorem . ) a ( countable ) borel equivalence relation @xmath133 is hyperfinite if @xmath1 is the increasing union of a sequence of finite borel equivalence relations on @xmath2 .
we will use the following facts ( see @xcite ) : [ thm : hyperfinite ] let @xmath153 be countable borel equivalence relations . 1 .
@xmath1 is hyperfinite iff @xmath223 for some action of @xmath20 on @xmath2 .
2 . @xmath1 is hyperfinite iff there is a borel binary relation @xmath224 on @xmath2 such that on each @xmath1-class , @xmath224 is a linear order embeddable in @xmath225 .
3 . if @xmath226 are both hyperfinite and non - smooth , then @xmath172 .
thus there is a unique bireducibility ( in fact biembeddability ) class of non - smooth hyperfinite borel equivalence relations .
let @xmath227 be the equivalence relations on @xmath228 given by @xmath229 up to isomorphism , the non - smooth , aperiodic , hyperfinite borel equivalence relations are @xmath230 where @xmath231 .
each @xmath232 has exactly @xmath104 ergodic invariant probability measures .
( glimm - effros dichotomy ) @xmath1 is not smooth iff @xmath212 .
a countable borel equivalence relation @xmath133 is compressible if there is a @xmath233 such that @xmath234 for every @xmath1-class @xmath235 .
the basic example is @xmath236 .
a fundamental theorem of nadkarni @xcite asserts that @xmath1 is compressible iff it does not admit an invariant probability measure .
thus , @xmath213 is compressible . for more on compressibility ,
see ( * ? ? ?
* section 2 ) ; we will use the results therein extensively in [ sec : compress ] .
a countable borel equivalence relation @xmath133 is treeable if @xmath1 is generated by an acyclic borel graph on @xmath2 . for properties of treeability which we use later on , see ( *
* section 3 ) .
let @xmath237 be borel equivalence relations , and let @xmath238 and @xmath239 be homomorphisms .
the fiber product of @xmath43 and @xmath240 ( with respect to @xmath47 and @xmath241 ) is @xmath242 , where @xmath243 the projections @xmath244 and @xmath245 fit into a commutative diagram : @xmath246{\pi_1 } \rar{\pi_2 } & g \dar{g } \\ f \rar[swap]{f } & e \end{tikzcd}\ ] ]
it is easily verified that if @xmath241 is class - injective , class - surjective , or a reduction , then so is @xmath247 . for each of the several kinds of homomorphisms mentioned in [ sec : homoms ] , we have a corresponding category of countable borel equivalence relations and homomorphisms of that kind .
we use e.g. , @xmath248 to denote the category of countable borel equivalence relations and class - bijective homomorphisms , etc .
( depending on context , we also use @xmath248 to denote the preorder @xmath187 on @xmath6 , i.e. , the preorder gotten by collapsing all morphisms in the category @xmath248 between the same two objects . ) from a categorical standpoint , among these categories , the two most well - behaved ones seem to be @xmath249 and @xmath248 .
the latter will be treated in [ sec : tensor , sec : limits ] .
as for @xmath249 , we note that ( countable ) disjoint sums , ( finite ) cross products , and fiber products give respectively coproducts , products , and pullbacks in that category .
it follows that @xmath249 is finitely complete , i.e. , has all finite categorical limits ( see e.g. , ( * ? ? ?
* v.2 , exercise iii.4.10 ) ) .
( however , it does not have coequalizers : let @xmath250 on @xmath228 be generated by a borel automorphism @xmath251 ; then it is easy to see that @xmath252 and the identity map do not have a coequalizer . ) for later reference , let us note that the category of ( not necessarily countable ) borel equivalence relations and borel homomorphisms has inverse limits of countable chains .
that is , for each @xmath253 , let @xmath254 be a borel equivalence relation , and @xmath255 be a borel homomorphism .
then the inverse limit of the system is @xmath256 , where @xmath257 it is easily seen that @xmath258 together with the projections @xmath259 has the universal property of an inverse limit , i.e. , for any other borel equivalence relation @xmath169 and homomorphisms @xmath260 such that @xmath261 for each @xmath262 , there is a unique homomorphism @xmath263 such that @xmath264 for each @xmath262 .
this is depicted in the following commutative diagram : @xmath265{\~g } \ar{dr}[swap , pos=.7,xshift=.5ex]{g_2 } \ar{drr}[swap , pos=.75,xshift=.5ex]{g_1 } \ar[bend left=5]{drrr}[swap , pos=.85,xshift=.5ex]{g_0 } & \projlim_n e_n \ar[crossing over]{d}[pos=.67,xshift=.3ex , swap]{\pi_2 } \ar[crossing over]{dr}[pos=.4,xshift=.7ex , swap]{\pi_1 } \ar[bend left=5]{drr}{\pi_0 } \\[2em ] \dotsb \rar[swap]{f_2 } & e_2 \rar[swap]{f_1 } & e_1 \rar[swap]{f_0 } & e_0 \end{tikzcd}\ ] ] it follows that the category of borel equivalence relations and borel homomorphisms is countably complete , i.e. , has all limits of countable diagrams ( again see ( * ? ? ? * v.2 , exercise iii.4.10 ) ) .
we now define the central notion of this paper .
let @xmath10 be a language and @xmath2 be a standard borel space .
we say that an @xmath10-structure @xmath110 with universe @xmath2 is borel if @xmath112 is borel for each @xmath104-ary @xmath111 .
now let @xmath133 be a countable borel equivalence relation .
we say that a borel @xmath10-structure @xmath110 is a borel @xmath10-structure on @xmath1 if for each @xmath104-ary @xmath111 , @xmath26 only relates elements within the same @xmath1-class , i.e. , @xmath266 for an @xmath33-sentence @xmath34 , we say that @xmath19 is a borel @xmath34-structure on @xmath1 , written @xmath267 if for each @xmath1-class @xmath235 , the structure @xmath17 satisfies @xmath34 .
we say that @xmath1 is @xmath34-structurable , written @xmath268 if there is some borel @xmath34-structure on @xmath1 .
similarly , if @xmath0 is a borel class of @xmath10-structures , we say that @xmath19 is a borel @xmath0-structure on @xmath1 if @xmath269 for each @xmath235 , and that @xmath1 is @xmath0-structurable if there is some borel @xmath0-structure on @xmath1 .
note that @xmath1 is @xmath0-structurable iff it is @xmath34-structurable , for any @xmath33-sentence @xmath34 axiomatizing @xmath0 .
we let @xmath270 denote respectively the classes of @xmath34-structurable and @xmath0-structurable countable borel equivalence relations . for any class @xmath66 of countable borel equivalence relations ,
we say that @xmath39 is elementary if @xmath271 for some theory @xmath109 , in which case we say that @xmath109 axiomatizes @xmath39 .
several notions of `` sufficiently simple '' countable borel equivalence relations which have been considered in the literature are given by an elementary class .
for example , a countable borel equivalence relation @xmath1 is smooth iff @xmath1 is structurable by pointed sets ( i.e. , sets with a distinguished element ) . by [ thm : hyperfinite ]
, @xmath1 is hyperfinite iff @xmath1 is structurable by linear orders which embed in @xmath20 .
hyperfiniteness can also be axiomatized by the sentence in the language @xmath272 which asserts that each @xmath8 is a finite equivalence relation and @xmath273 with union the indiscrete equivalence relation .
similarly , it is straightforward to verify that for each @xmath274 , @xmath275-frchet - amenability ( see @xcite ) is axiomatizable . also , @xmath1 is compressible iff it is structurable via structures in the language @xmath276 where @xmath23 is the graph of a non - surjective injection . for some trivial examples :
every @xmath1 is @xmath34-structurable for logically valid @xmath34 , or for the ( non - valid ) sentence @xmath34 in the language @xmath277 asserting that the @xmath8 s form a separating family of unary predicates ( i.e. , @xmath278 ) ; thus @xmath6 is elementary . the class of aperiodic countable borel equivalence is axiomatized by the theory of infinite sets , etc .
let @xmath279 denote the class of trees ( i.e. , acyclic connected graphs ) , and more generally , @xmath280 denote the class of contractible @xmath104-dimensional ( abstract ) simplicial complexes .
then @xmath1 is @xmath279-structurable iff @xmath1 is treeable .
gaboriau @xcite has shown that @xmath281 .
for any language @xmath10 and countable @xmath10-structure @xmath19 , if @xmath93 denotes the scott sentence of @xmath19 , then @xmath1 is @xmath93-structurable iff it is structurable via isomorphic copies of @xmath19 .
for example , if @xmath282 and @xmath283 , then @xmath1 is @xmath93-structurable iff it is aperiodic hyperfinite .
we write @xmath284 for the class of @xmath19-structurable countable borel equivalence relations .
let @xmath21 be a countable group , and regard @xmath21 as a structure in the language @xmath285 , where @xmath286 is the graph of the map @xmath287 .
then a model of @xmath288 is a @xmath21-action isomorphic to @xmath21 , i.e. , a free transitive @xmath21-action .
thus a countable borel equivalence relation @xmath1 is @xmath21-structurable ( i.e. , @xmath288-structurable ) iff it is generated by a free borel action of @xmath21 .
finally , we note that several important classes of countable borel equivalence relations are _ not _ elementary .
this includes all classes of `` sufficiently complex '' equivalence relations , such as ( invariantly ) universal equivalence relations , non - smooth equivalence relations , and equivalence relations admitting an invariant probability measure ; these classes are not elementary by [ thm : elem - classbij ] .
another example of a different flavor is the class of equivalence relations generated by a free action of _ some _ countable group ; more generally , nontrivial unions of elementary classes are never elementary ( see [ thm : union - nonelem ] ) .
let @xmath190 be countable borel equivalence relations and @xmath289 be a class - bijective homomorphism .
for an @xmath10-structure @xmath19 on @xmath43 , recall that the pullback structure of @xmath19 along @xmath47 , denoted @xmath125 , is the @xmath10-structure with universe @xmath2 given by @xmath290 for each @xmath104-ary @xmath111 and @xmath291 .
let @xmath292 denote the classwise pullback structure , given by @xmath293 then @xmath292 is a borel @xmath10-structure on @xmath1 , such that for each @xmath1-class @xmath235 , the restriction @xmath294 is an isomorphism between @xmath295 and @xmath296 . in particular , if @xmath19 is a @xmath34-structure for some @xmath33-sentence @xmath34 , then so is @xmath292 .
we record the consequence of this simple observation for structurability : [ thm : elem - classbij ] every elementary class @xmath297 is ( downwards-)closed under class - bijective homomorphisms , i.e. , if @xmath181 and @xmath298 , then @xmath299 .
this connection between structurability and class - bijective homomorphisms will be significantly strengthened in the next section .
in this section we present the two main constructions relating structures on equivalence relations to class - bijective homomorphisms .
both are `` universal '' constructions : the first turns any theory @xmath109 into a universal equivalence relation with a @xmath34-structure , while the second turns any equivalence relation into a universal theory .
kechris - solecki - todorcevic @xcite proved a universality result for graphs , which was then extended to arbitrary borel classes of structures by miller . here , we formulate a version of this result and its proof that allows us to capture more information .
[ thm : esigma ] let @xmath300 be a countable borel equivalence relation and @xmath109 be a theory .
then there is a `` universal @xmath34-structured equivalence relation lying over @xmath1 '' , i.e. , a triple @xmath301 where @xmath302 such that for any other @xmath303 with @xmath304 and @xmath305 , there is a _
unique _ class - bijective homomorphism @xmath306 such that @xmath307 and @xmath308 .
this is illustrated by the following `` commutative '' diagram : @xmath309{ddr}[swap]{f } \drar[dashed][pos=.7,swap]{\~f } \ar[models]{drr}{\#a } & & [ -.5em ] \\ & e \ltimes \sigma
\dar{\pi } \rar[models][swap]{\#e } & \sigma \\ & e \end{tikzcd}\ ] ] first we describe @xmath310 while ignoring all questions of borelness , then we verify that the construction can be made borel . ignoring borelness
, @xmath310 will live on a set @xmath311 and will have the following form : for each @xmath1-class @xmath235 , and each @xmath34-structure @xmath312 on the universe @xmath16 , there will be one @xmath313-class lying over @xmath16 ( i.e. , projecting to @xmath16 via @xmath314 ) , which will have the @xmath34-structure given by pulling back @xmath312 .
thus we put @xmath315_e}(\sigma)\ } , \\ ( x , \#b ) \mathrel{(e \ltimes \sigma ) } ( x ' , \#b ' ) \iff x \mathrel{e } x ' \and \#b = \#b ' , \\
\pi(x , \#b ) : = x,\end{gathered}\ ] ] with the @xmath34-structure @xmath316 on @xmath310 given by @xmath317 for @xmath104-ary @xmath111 , @xmath318 , and @xmath319_e}(\sigma)$ ] .
it is immediate that @xmath314 is class - bijective and that @xmath316 satisfies @xmath34 .
the universal property is also straightforward : given @xmath320 as above , the map @xmath321 is given by @xmath322_f ) ) \in z,\end{aligned}\ ] ] and this choice is easily seen to be unique by the requirements @xmath307 and @xmath308 .
now we indicate how to make this construction borel .
the only obstruction is the use of @xmath323_e}(\sigma)$ ] which depends on @xmath164 in the definition of @xmath311 above .
we restrict to the case where @xmath1 is aperiodic ; in general , we may split @xmath1 into its finite part and aperiodic part , and it will be clear that the finite case can be handled similarly . in the aperiodic case ,
the idea is to replace @xmath323_e}(\sigma)$ ] with @xmath324 , where @xmath4_e$ ] is identified with @xmath325 but in a manner which varies depending on @xmath164 .
let @xmath326 be a borel map such that each @xmath327 is a bijection @xmath328_e$ ] ( the existence of such @xmath329 is easily seen from lusin - novikov uniformization ) , and replace @xmath323_e}(\sigma)$ ] with @xmath324 while inserting @xmath327 into the appropriate places in the above definitions : @xmath330_f))).\end{gathered}\ ] ] these are easily seen to be borel and still satisfy the requirements of the theorem .
it is clear that @xmath310 satisfies a universal property in the formal sense of category theory .
this in particular means that @xmath301 is unique up to _ unique _ ( borel ) isomorphism .
[ rmk : skew - product ] the construction of @xmath310 for aperiodic @xmath1 in the proof of [ thm : esigma ] can be seen as an instance of the following general notion ( see e.g. , ( * ? ? ?
* e ) ) : let @xmath133 be a borel equivalence relation , and let @xmath21 be a ( borel ) group . recall that a borel cocycle @xmath331 is a borel map satisfying @xmath332 for all @xmath333 , @xmath334 . given a cocycle @xmath275 and a borel action of @xmath21 on a standard borel space @xmath118 , the skew product @xmath335 is the borel equivalence relation on @xmath336 given by @xmath337 note that for such a skew product , the first projection @xmath200 is always a class - bijective homomorphism @xmath338 . now given a family @xmath326 of bijections @xmath339_e$ ] , as in the proof of [ thm : esigma ]
, we call @xmath340 given by @xmath341 the cocycle induced by @xmath329
. then the construction of @xmath310 for aperiodic @xmath1 can be seen as the skew product @xmath342 ( with the logic action of @xmath343 on @xmath324 ) .
( however , the structure @xmath316 on @xmath1 depends on @xmath329 , not just on @xmath275 . )
[ thm : esigma ] has the following consequence : [ thm : einftysigma ] for every theory @xmath109 , there is an invariantly universal @xmath34-structurable countable borel equivalence relation @xmath62 , i.e. , @xmath63 , and @xmath64 for any other @xmath65 .
put @xmath344 .
for any @xmath65 , we have an invariant embedding @xmath345 , whence there is @xmath346 such that @xmath307 ; since @xmath47 is injective , so is @xmath321 . in other words ,
every elementary class @xmath40 of countable borel equivalence relations has an invariantly universal element @xmath62 ( which is unique up to isomorphism ) . for a borel class of structures @xmath0
, we denote the invariantly universal @xmath0-structurable equivalence relation by @xmath347 . for an @xmath10-structure @xmath19
, we denote the invariantly universal @xmath19-structurable equivalence relation by @xmath348 . as a basic application , we can now rule out the elementarity of a class of equivalence relations mentioned in [ sec : elem - examples ] : [ thm : union - nonelem ] if @xmath349 is a collection of elementary classes of countable borel equivalence relations , then @xmath350 is _ not _ elementary , unless there is some @xmath351 such that @xmath352 for every @xmath353 .
in particular , the class of equivalence relations generated by a free borel action of some countable group is _ not _ elementary .
if @xmath350 is elementary , then it has an invariantly universal element @xmath1 , which is in some @xmath354 ; then for every @xmath353 and @xmath355 , we have @xmath356 , whence @xmath357 since @xmath354 is elementary . for the second statement , the class in question is @xmath358 where @xmath21 ranges over countable groups ( and @xmath76 is the class of equivalence relations generated by a free borel action of @xmath21 ) ; and there can not be a single @xmath76 which contains all others , since if @xmath21 is amenable then @xmath76 does not contain @xmath359 ( see ( * ? ? ? *
a4.1 ) ) , while if @xmath21 is not amenable then @xmath76 does not contain @xmath250 ( see @xcite ) .
we also have let @xmath39 denote the class of countable increasing unions of equivalence relations generated by free borel actions of ( possibly different ) countable groups .
then @xmath39 does not have a @xmath86-universal element , hence is not elementary .
let @xmath360 be the countable increasing union of countable borel equivalence relations @xmath361 on @xmath2 , where each @xmath362 is generated by a free borel action of a countable group @xmath363 .
since there are uncountably many finitely generated groups , there is a finitely generated group @xmath10 such that @xmath10 does not embed in any @xmath363 .
put @xmath364 , and let @xmath365 live on @xmath366 ( the free part of the shift action ) , with its usual product probability measure @xmath147 .
by @xcite ( see also 3.79 of that paper ) , @xmath367 for each @xmath104 and @xmath368 of @xmath147-measure @xmath369 .
if @xmath1 were @xmath86-universal in @xmath39 , then we would have some @xmath370 .
let @xmath371 , so that @xmath372 . by @xcite
( using that @xmath373 acts strongly ergodically ( * ? ? ?
* a4.1 ) ) , there is an @xmath104 and a borel @xmath374 with @xmath375 such that @xmath376 . by ergodicity of @xmath147 , @xmath377_{f(\delta , 2)}$ ]
has @xmath147-measure @xmath369 ; but @xmath378 , a contradiction .
we conclude this section by explicitly describing the invariantly universal equivalence relation in several elementary classes : * the @xmath220-universal finite borel equivalence relation is @xmath379 . *
the @xmath220-universal aperiodic smooth countable borel equivalence relation is @xmath380 . *
the @xmath220-universal aperiodic hyperfinite borel equivalence relation is @xmath381 , and the @xmath220-universal compressible hyperfinite borel equivalence relation is @xmath213 ( see [ thm : hyperfinite ] ) . *
the @xmath220-universal countable borel equivalence relation is @xmath221 , and the @xmath220-universal compressible borel equivalence relation is @xmath382 ( see [ sec : compress ] ) .
* for a countable group @xmath21 , the @xmath220-universal equivalence relation @xmath383 generated by a free borel action of @xmath21 is @xmath384 .
we now associate to every @xmath69 a `` scott sentence '' @xmath385 . just as the scott sentence @xmath93 of an ordinary first - order structure @xmath19 axiomatizes structures isomorphic to @xmath19 , the `` scott sentence '' @xmath385 will axiomatize equivalence relations class - bijectively mapping to @xmath1 .
[ thm : sigmae ] let @xmath300 be a countable borel equivalence relation . then there is a sentence @xmath385 ( in some fixed language not depending on @xmath1 ) and a @xmath385-structure @xmath386 , such that for any @xmath303 and @xmath387 , there is a unique class - bijective homomorphism @xmath304 such that @xmath388 .
this is illustrated by the following diagram : @xmath389[swap]{f } \drar[models]{\#a } \\ e \rar[models][swap]{\#h } & \sigma_e \end{tikzcd}\ ] ] we may assume that @xmath2 is a borel subspace of @xmath228 .
let @xmath277 where each @xmath8 is unary .
the idea is that a borel @xmath10-structure will code a borel map to @xmath390 .
note that since @xmath10 is unary , there is no distinction between borel @xmath10-structures on @xmath2 and borel @xmath10-structures on @xmath1 , or between pullback @xmath10-structures and classwise pullback @xmath10-structures .
let @xmath391 be the borel @xmath10-structure on @xmath228 given by @xmath392 it is clear that for any standard borel space @xmath118 , we have a bijection @xmath393 it will suffice to find an @xmath33-sentence @xmath385 such that for all @xmath394 and @xmath395 , @xmath396 indeed , we may then put @xmath397 , and ( @xmath398 ) will restrict to a bijection between class - bijective homomorphisms @xmath182 and @xmath385-structures on @xmath43 , as claimed in the theorem .
now we find @xmath385 satisfying ( @xmath399 ) .
the conditions @xmath400 and @xmath304 can be rephrased as : for each @xmath43-class @xmath401 , the restriction @xmath402 is a bijection between @xmath403 and some @xmath1-class . using ( @xmath398 ) , this is equivalent to : for each @xmath43-class @xmath401 , the structure @xmath404 on @xmath403 is such that @xmath405 so it suffices to show that the class @xmath0 of @xmath10-structures @xmath312 satisfying ( @xmath406 ) is borel ( so we may let @xmath385 be any @xmath33-sentence axiomatizing @xmath0 ) , i.e. , that for any @xmath407 , @xmath408 is borel . using ( @xmath398 ) again , @xmath409 is the image of the borel injection @xmath410 the domain of this injection is clearly a borel subset of @xmath411 , whence its image is borel . in the rest of this section
, we give an alternative , more `` explicit '' construction of @xmath385 ( rather than obtaining it from lopez - escobar s definability theorem as in the above proof ) .
using the same notations as in the proof , we want to find @xmath385 satisfying ( @xmath399 ) . by lusin - novikov uniformization ,
write @xmath412 where @xmath413 are graphs of ( total ) borel functions . for each @xmath353
, let @xmath414 be a quantifier - free @xmath33-formula whose interpretation in the structure @xmath391 is @xmath415 .
( such a formula can be obtained from a borel definition of @xmath416 in terms of the basic rectangles @xmath417 , by replacing each @xmath417 with @xmath418 . )
define the @xmath33-sentences : @xmath419 [ lm : sigmae ] in the notation of the proof of [ thm : sigmae ] , @xmath420 @xmath421 iff for all @xmath422 , there is some @xmath353 such that @xmath423 ; @xmath423 is equivalent to @xmath424 , i.e. , @xmath425 , so we get that @xmath421 iff for all @xmath422 , we have @xmath426 . ( taking @xmath427 yields @xmath428 . )
@xmath429 iff for all @xmath422 with @xmath430 , there is some @xmath353 such that @xmath431 , i.e. , @xmath432 , i.e. , @xmath433 .
@xmath434 iff for all @xmath435 and all @xmath436 , there is some @xmath437 such that @xmath423 , i.e. , @xmath424 , i.e. , @xmath425 ; from the definition of the @xmath438 , this is equivalent to : for all @xmath435 , we have @xmath428 , and for every @xmath439 there is some @xmath437 such that @xmath440 . so defining @xmath441 , we have that @xmath442 iff @xmath304 , as desired .
moreover , by modifying these sentences , we may obtain theories for which structures on @xmath43 correspond to other kinds of homomorphisms @xmath443 .
we will take advantage of this later , in [ sec : classinj , sec : smh ] .
the combination of [ thm : esigma , thm : sigmae ] gives the following ( closely related ) corollaries , which imply a tight connection between structurability and class - bijective homomorphisms .
[ thm : classbij - elem ] for @xmath87 , we have @xmath444 iff @xmath182 . by [ thm : sigmae , thm : elem - classbij ] . [
thm : elem - cone ] for every @xmath69 , there is a smallest elementary class containing @xmath1 , namely @xmath445 . by [ thm : elem - classbij ] , @xmath446 is contained in every elementary class containing @xmath1 .
we define @xmath447 , and call it the elementary class of @xmath1 .
[ thm : elem - char ] a class @xmath66 of countable borel equivalence relations is elementary iff it is ( downwards-)closed under class - bijective homomorphisms and contains an invariantly universal element @xmath67 , in which case @xmath448 . one implication is [ thm : elem - classbij , thm : einftysigma ] .
conversely , if @xmath39 is closed under @xmath187 and @xmath67 is invariantly universal , then clearly @xmath449 .
so every elementary class @xmath39 is determined by a canonical isomorphism class contained in @xmath39 , namely the invariantly universal elements of @xmath39 .
we now characterize the class of equivalence relations which are invariantly universal in some elementary class .
[ thm : univstr ] let @xmath69 .
the following are equivalent : 1 .
@xmath450 , i.e. , @xmath1 is invariantly universal in @xmath451 .
2 . @xmath452 for some @xmath34 , i.e. , @xmath1 is invariantly universal in some elementary class .
3 . for every @xmath303 , @xmath182 iff @xmath185 . clearly
( i)@xmath453(ii)@xmath453(iii ) , and if ( iii ) holds , then @xmath454 so @xmath1 is invariantly universal in @xmath451 .
the awkward notation @xmath455 will be replaced in the next section ( with @xmath456 ) .
we say that @xmath69 is universally structurable if the equivalent conditions in [ thm : univstr ] hold .
we let @xmath457 denote the class of universally structurable countable borel equivalence relations .
the following summarizes the relationship between @xmath458 and elementary classes : [ thm : univstr - elem ] we have an order - isomorphism of posets @xmath459 we will study the purely order - theoretical aspects of the poset @xmath82 ( equivalently , the poset of elementary classes ) in [ sec : poset ] .
we conclude this section by pointing out the following consequence of universal structurability : [ thm : univstr - idemp ] if @xmath69 is universally structurable , then @xmath460 . in particular
, @xmath1 has either none or continuum many ergodic invariant probability measures . clearly @xmath461 , and @xmath462 , so @xmath463 . in this section and the next , we use the theory of the preceding sections to obtain some structural results about the category @xmath248 of countable borel equivalence relations and class - bijective homomorphisms . for categorical background , see @xcite .
this section concerns a certain product construction between countable borel equivalence relations , which , unlike the cross product @xmath464 , is well - behaved with respect to class - bijective homomorphisms .
[ thm : tensor ] let @xmath87 be countable borel equivalence relations .
there is a countable borel equivalence relation , which we denote by @xmath465 and call the class - bijective product ( or tensor product ) of @xmath1 and @xmath43 , which is the categorical product of @xmath1 and @xmath43 in the category @xmath248 .
in other words , there are canonical class - bijective projections @xmath466 such that the triple @xmath467 is universal with these properties : for any other @xmath468 with @xmath469 and @xmath470 , there is a _
unique _ class - bijective homomorphism @xmath471 such that @xmath472 and @xmath473 .
this is illustrated by the following commutative diagram : @xmath474{f } \drar{g } \dar[dashed][pos=.7]{\ang{f , g } } \\ e & e \otimes f \lar{\pi_1 } \rar[swap]{\pi_2 } & f \end{tikzcd}\ ] ] put @xmath475 .
the rest follows from chasing through the universal properties in [ thm : esigma ] and [ thm : sigmae ] ( or equivalently , the yoneda lemma ) . for the sake of completeness , we give the details . from [ thm : esigma ] , we have a canonical projection @xmath476 .
we also have a canonical @xmath477-structure on @xmath465 , namely @xmath478 .
this structure corresponds to a unique class - bijective map @xmath479 such that @xmath480 , where @xmath481 is the canonical structure from [ thm : sigmae ] .
now given @xmath482 as above , the map @xmath483 is produced as follows .
we have the classwise pullback structure @xmath484 , which , together with @xmath469 , yields ( by [ thm : esigma ] ) a unique map @xmath485 such that @xmath472 and @xmath486 . since @xmath480 , we get @xmath487 ; since ( by [ thm : sigmae ] ) @xmath241 is the _ unique _
map @xmath488 such that @xmath489 , we get @xmath473 , as desired .
it remains to check uniqueness of @xmath483 .
if @xmath490 is such that @xmath491 and @xmath492 , then ( reversing the above steps ) we have @xmath493 ; since @xmath483 was unique with these properties , we get @xmath494 , as desired . as with all categorical products ,
@xmath495 is unique up to unique ( borel ) isomorphism , as well as associative and commutative up to ( borel ) isomorphism .
note that the two latter properties are not immediately obvious from the definition @xmath475 .
[ thm : tensor - alt ] however , by unravelling the proofs of [ thm : esigma , thm : sigmae ] , we may explicitly describe @xmath465 in a way that makes associativity and commutativity more obvious .
since this explicit description also sheds some light on the structure of @xmath465 , we briefly give it here .
let @xmath1 live on @xmath2 , @xmath43 live on @xmath118 , and @xmath465 live on @xmath311 .
we have one @xmath496-class for each @xmath1-class @xmath16 , @xmath43-class @xmath403 , and bijection @xmath497 ; the elements of the @xmath496-class corresponding to @xmath498 are the elements of @xmath16 , or equivalently via the bijection @xmath499 , the elements of @xmath403 .
thus , ignoring borelness , we put @xmath500_e \cong [ y]_f,\ , b(x ) = y\ } , \\ ( x , y , b ) \mathrel{(e \otimes f ) } ( x ' , y ' , b ' ) \iff x \mathrel{e } x ' \and y \mathrel{f } y ' \and b = b ' , \\ \pi_1(x , y , b ) : = x , \hspace{1 in } \pi_2(x , y , b ) : = y.\end{gathered}\ ] ] given @xmath482 as in [ thm : tensor ] ( @xmath240 living on @xmath501 , say ) , the map @xmath471 is given by @xmath502_g ) \circ ( f|[w]_g)^{-1})$ ] , where @xmath503_g ) \circ ( f|[w]_g)^{-1 } ) : [ f(w)]_e \cong [ g(w)]_f$ ] since @xmath504 are class - bijective . to make this construction borel
, we assume that @xmath226 are aperiodic , and replace @xmath311 in the above with a subspace of @xmath505 , where bijections @xmath506_e \cong [ y]_f$ ] are transported to bijections @xmath507 via borel enumerations of the @xmath1-classes and @xmath43-classes , as with the map @xmath326 in the proof of [ thm : esigma ] .
( if @xmath226 are not aperiodic , then we split them into the parts consisting of classes with each cardinality @xmath508 ; then there will be no @xmath496-classes lying over an @xmath1-class and an @xmath43-class with different cardinalities . ) the tensor product @xmath495 and the cross product @xmath509 are related as follows : we have a canonical homomorphism @xmath510 , where @xmath511 ( where @xmath476 and @xmath479 are the projections from the tensor product ) , which is class - injective because @xmath512 is class - injective ( where @xmath513 is the projection from the _ cross _ product ) .
when we regard @xmath465 as in [ thm : tensor - alt ] , @xmath514 is the obvious projection from @xmath311 to @xmath336 .
this in particular shows that [ thm : tensor - cross ] 1 .
@xmath510 is surjective iff @xmath1 and @xmath43 have all classes of the same cardinality ( in particular , if both are aperiodic ) ; 2 .
@xmath510 is an isomorphism if @xmath515 and @xmath516 .
we now list some formal properties of @xmath495 : [ thm : tensor - props ] let @xmath517 for @xmath518 and let @xmath109 be a theory . 1 .
if @xmath519 then @xmath520 .
2 . if @xmath521 and @xmath522 , then @xmath523 .
3 . if @xmath1 is universally structurable , then so is @xmath465 .
4 . @xmath524 ( and the isomorphism is natural in @xmath226 ) .
@xmath525 ( and the isomorphism is natural in @xmath526 ) .
( a ) : follows from @xmath476 .
( b ) : since @xmath472 is class - injective , so is @xmath483 .
( c ) : if @xmath527 , then @xmath528 , whence there is some @xmath529 since @xmath1 is universally structurable , whence @xmath530 ( by ( b ) ) .
( d ) : follows from a chase through the universal properties of @xmath495 and @xmath531 ( or the yoneda lemma ) .
( a class - bijective homomorphism @xmath532 is the same thing as pair of class - bijective homomorphisms @xmath533 and @xmath534 together with a @xmath34-structure on @xmath240 , which is the same thing as a class - bijective homomorphism @xmath535 . ) ( e ) : this is an instance of the following more general fact , which follows easily from the construction of @xmath310 in [ thm : esigma ] ( and which could have been noted earlier , in [ sec : esigma ] ) : [ thm : esigma - dist ] @xmath536 .
moreover , the isomorphism can be taken to be the map @xmath537 such that for each @xmath353 , the restriction @xmath538 is the canonical such map induced by the inclusion @xmath539 .
( in other words , the functor @xmath540 preserves countable coproducts . ) to get from this to ( e ) , simply put @xmath541 .
naturality is straightforward .
the analog of [ thm : tensor - props](a ) for cross products is false : the class of treeable countable borel equivalence relations is _ not _ closed under cross products ( see e.g. , @xcite ) .
we note that for any @xmath69 , the equivalence relation @xmath455 ( i.e. , the invariantly universal element of @xmath451 ) can also be written as the less awkward @xmath456 , which is therefore how we will write it from now on . here
are some sample computations of class - bijective products : * @xmath542 for @xmath543 ( by [ thm : tensor - cross](b ) ) .
* @xmath544 , since @xmath545 is aperiodic smooth ( [ thm : tensor - props](a ) ) and there are continuum many bijections @xmath507 .
* @xmath546 , since @xmath547 ( [ thm : tensor - props](b ) ) . * if @xmath1 is universally structurable and @xmath181 , then @xmath548 , since @xmath549 ( [ thm : tensor - props](b ) ) , and @xmath476 so @xmath550 . * @xmath551 , since @xmath552 is aperiodic hyperfinite ( [ thm : tensor - props](a ) ) , and there are @xmath553 pairwise disjoint copies of @xmath250 in @xmath552 .
this last fact can be seen by taking a family @xmath554 of borel automorphisms @xmath555 such that @xmath556 disagree on every @xmath250-class whenever @xmath557 , and then considering the embeddings @xmath558 , which will have pairwise disjoint images ( by [ thm : tensor - alt ] ) .
+ ( the existence of the family @xmath559 is standard ; one construction is by regarding @xmath250 as the orbit equivalence of the translation action of @xmath20 on @xmath560 , the @xmath219-adic integers , then taking @xmath561 for @xmath562 to be translation by @xmath563 . ) note that the last example can be used to compute @xmath495 of all hyperfinite equivalence relations .
we conclude by noting that we do not currently have a clear picture of tensor products of general countable borel equivalence relations .
for instance , the examples above suggest that perhaps @xmath564 is universally structurable ( equivalently , @xmath565 , since @xmath566 ) for all aperiodic @xmath1 ; but we do not know if this is true . this short section concerns general categorical limits in the category @xmath248 of countable borel equivalence relations and class - bijective homomorphisms . throughout this section ,
we use categorical terminology , e.g. , `` product '' means categorical product ( i.e. , class - bijective product ) , `` pullback '' means categorical pullback ( i.e. , fiber product ) , etc . for definitions ,
see ( * ? ? ?
* iii.34 , v ) .
we have shown that @xmath248 contains binary products . by iterating ( or by generalizing the construction outlined in [ thm : tensor - alt ] )
, we may obtain all finite ( nontrivial ) products .
the category @xmath248 also contains pullbacks ( see [ sec : pullback ] ) .
it follows that it contains all finite nonempty limits , i.e. , limits of all diagrams @xmath567 where the indexing category @xmath568 is finite and nonempty ( see ( * ? ? ?
* v.2 , exercise iii.4.9 ) ) .
@xmath248 does _ not _ contain a terminal object , i.e. , a limit of the empty diagram .
this would be a countable borel equivalence relation @xmath1 such that any other countable borel equivalence relation @xmath43 has a _
unique _ class - bijective map @xmath182 ; clearly such @xmath1 does not exist .
we now verify that @xmath248 has inverse limits of countable chains , and that these coincide with the same limits in the category of all borel equivalence relations and borel homomorphisms ( [ sec : prelims - cats ] ) : let @xmath569 be countable borel equivalence relations , and @xmath570 be class - bijective homomorphisms .
then the inverse limit @xmath571 in the category of borel equivalence relations and borel homomorphisms , as defined in [ sec : prelims - cats ] , is also the inverse limit in @xmath248 .
more explicitly , 1 .
the projections @xmath259 are class - bijective ( so in particular @xmath258 is countable ) ; 2 . if @xmath394 is a countable borel equivalence relation with class - bijective homomorphisms @xmath572 such that @xmath573 , then the unique homomorphism @xmath263 such that @xmath574 for each @xmath104 ( namely @xmath575 ) is class - bijective . for ( a ) , note that since @xmath576 and the @xmath577 are class - bijective , it suffices to check that @xmath578 is class - bijective , since we may then inductively get that @xmath579 are class - bijective .
let @xmath580 and @xmath581 ; we must find a unique @xmath582 such that @xmath583 . for the coordinate @xmath584 , we must have @xmath585 ( in order to have @xmath586 ) and @xmath587 ( in order to have @xmath582 ) ; since @xmath588 , by class - bijectivity of @xmath589 , there is a unique such @xmath590 .
continuing inductively , we see that there is a unique choice of @xmath591 for each @xmath592 .
then @xmath593 is the desired element . for ( b ) , simply note that since @xmath574 and @xmath594 are class - bijective , so must be @xmath595 .
@xmath248 has all countable ( nontrivial ) products . to compute the product @xmath596 of @xmath597 ,
take the inverse limit of the chain @xmath598 ( where the maps are the projections ) .
countable products can also be obtained by generalizing [ thm : tensor - alt ] .
@xmath248 has all countable nonempty limits , i.e. , limits of all diagrams @xmath567 , where the indexing category @xmath568 is countable and nonempty
. follows from countable products and pullbacks ; again see ( * ? ? ?
* v.2 , exercise iii.4.9 ) .
this section has two parts : the first part ( ) relates structurability to various classes of homomorphisms , in the spirit of ; while the second part ( [ sec : compress ] ) relates reductions to compressibility , using results from the first part and from ( * ? ? ?
* section 2 ) .
we describe here the various classes of homomorphisms that we will be considering .
these fit into the following table : .global and local classes of homomorphisms [ cols="^,^",options="header " , ] the last entry in the table denotes the following notion : we say that a borel homomorphism @xmath155 is smooth , written @xmath599 , if the @xmath47-preimage of every smooth set is smooth ( where by a smooth subset of @xmath118 ( resp . ,
@xmath2 ) we mean a subset to which the restriction of @xmath43 ( resp . ,
@xmath1 ) is smooth ) .
this notion was previously considered by clemens - conley - miller @xcite , under the name smooth - to - one homomorphism ( because of [ thm : smh - factor](ii ) ) .
see [ thm : smh - factor ] for basic properties of smooth homomorphisms .
let @xmath600 be a row in [ tbl : global - local ] .
we say that @xmath601 is a `` global '' class of homomorphisms , while @xmath602 is the corresponding `` local '' class .
note that @xmath603 .
the idea is that @xmath601 is a condition on homomorphisms requiring injectivity between classes ( i.e. , @xmath601 consists only of reductions ) , while @xmath602 is an analogous `` classwise '' condition which can be captured by structurability . our main results in this section state the following : for any elementary class @xmath66 , the downward closure of @xmath39 under @xmath601 is equal to the downward closure under @xmath602 , and is elementary . in particular ,
when @xmath448 , this implies that the downward closure of @xmath604 under @xmath602 is elementary . in the case
@xmath605 , these follow from [ sec : univstr ] ; thus , our results here generalize our results therein to the other classes of homomorphisms appearing in [ tbl : global - local ] .
[ thm : emb - elem ] let @xmath66 be an elementary class
. then the downward closures of @xmath39 under @xmath186 and @xmath606 , namely @xmath607 are equal and elementary . in particular , when @xmath448 , we get that @xmath608 is the smallest elementary class containing @xmath1 and closed under @xmath186 .
[ thm : red - elem ] let @xmath66 be an elementary class
. then the downward closures of @xmath39 under @xmath86 and @xmath609 , namely @xmath610 are equal and elementary .
in particular , when @xmath448 , we get that @xmath611 is the smallest elementary class containing @xmath1 and closed under @xmath86 .
our proof strategy is as follows . for each @xmath600 (= @xmath612 or @xmath613 ) , we prove a `` factorization lemma '' which states that @xmath602 consists precisely of composites of homomorphisms in @xmath601 followed by class - bijective homomorphisms ( in that order ) .
this yields that the closures of @xmath39 under @xmath601 and @xmath602 are equal , since @xmath39 is already closed under class - bijective homomorphisms .
we then prove that for any @xmath69 , a variation of the `` scott sentence '' from [ thm : sigmae ] can be used to code @xmath602-homomorphisms to @xmath1 .
this yields that for any @xmath69 , the @xmath602-downward closure of @xmath604 is elementary , which completes the proof .
we begin with embeddings and class - injective homomorphisms , for which we have the following factorization lemma : [ thm : classinj - factor ] let @xmath153 be countable borel equivalence relations and @xmath614 be a class - injective homomorphism . then there is a countable borel equivalence relation @xmath615 , an embedding @xmath616 , and a class - bijective homomorphism @xmath488 , such that @xmath617 : @xmath618 ( x , e ) \drar[dashed][swap]{g } \ar{rr}{f } & & ( y , f ) \\ & ( z , g ) \urar[dashed][swap]{h } \end{tikzcd}\ ] ] furthermore
, @xmath241 can be taken to be a complete section embedding , i.e. , @xmath619_g = z$ ] .
consider the equivalence relation @xmath620 where @xmath621 then @xmath620 is a countable borel equivalence relation .
we claim that it is smooth .
indeed , by lusin - novikov uniformization , write @xmath622 where @xmath623 for @xmath436 are graphs of borel functions @xmath624 .
then a borel selector for @xmath403 is found by sending @xmath625 to @xmath626_e)^{-1}(g_i(y ) ) , y)$ ] for the least @xmath353 such that @xmath627 is in the image of @xmath48_e$ ] .
( here we are using that @xmath47 is class - injective . )
now put @xmath628 , and let @xmath240 be the equivalence relation on @xmath311 given by @xmath629_d \mathrel{g } [ ( x ' , y')]_d \iff x \mathrel{e } x ' \and y \mathrel{f } y'.\end{aligned}\ ] ] then @xmath630 is a countable borel equivalence relation .
let @xmath631 and @xmath632 be given by @xmath633_d , & h([(x , y)]_d ) & : = y.\end{aligned}\ ] ] it is easily seen that @xmath616 is a complete section embedding , @xmath488 , and @xmath617 , as desired . it is easy to see that the factorization produced by [ thm : classinj - factor ] ( with the requirement that @xmath241 be a complete section embedding ) is unique up to unique borel isomorphism .
in other words , if @xmath634 , @xmath635 , and @xmath636 with @xmath637 are another factorization , with @xmath638 a complete section embedding , then there is a unique borel isomorphism @xmath639 such that @xmath640 and @xmath641 .
[ thm : univstr - classinj - emb ] if @xmath69 is universally structurable , then @xmath642 , for all @xmath303 .
similarly , if @xmath39 is an elementary class , then @xmath643 .
if @xmath1 is universally structurable and @xmath644 , then by [ thm : classinj - factor ] , @xmath645 for some @xmath240 ; then @xmath646 , whence @xmath647 .
the second statement is similar .
we now have the following analog of [ thm : sigmae ] for class - injective homomorphisms , which we state in the simpler but slightly weaker form of [ thm : classbij - elem ] since that is all we will need : [ thm : classinj - elem ] let @xmath69 be a countable borel equivalence relation . then there is a sentence @xmath648 ( in some fixed language ) such that for all @xmath303 , we have @xmath649 iff @xmath644 .
we may either modify the proof of [ thm : sigmae ] ( by considering `` injections @xmath650 ( some @xmath1-class ) '' instead of bijections in the last few lines of the proof ) , or take @xmath651 where @xmath652 and @xmath653 are as in [ lm : sigmae ] .
if @xmath448 is an elementary class , then so is @xmath654 .
this completes the proof of [ thm : emb - elem ] .
recall that a borel homomorphism @xmath655 between countable borel equivalence relations @xmath226 is smooth if the preimage of every smooth set is smooth .
we have the following equivalent characterizations of smooth homomorphisms , parts of which are implicit in @xcite : [ thm : smh - factor ] let @xmath153 and @xmath655 .
the following are equivalent : 1 .
@xmath47 is smooth .
2 . for every @xmath435
, @xmath656 is smooth ( i.e. , @xmath657 is smooth ) .
3 . @xmath658 is smooth ( as a countable borel equivalence relation on @xmath2 ) .
@xmath47 can be factored into a surjective reduction @xmath659 , followed by a complete section embedding @xmath660 , followed by a class - bijective homomorphism @xmath661 , for some @xmath662 : @xmath618 e \drar[dashed][swap]{g } \ar{rrr}{f } & & [ 1ex ] & f \\ & g \rar[dashed][swap]{h } & h \urar[dashed][swap]{k } \end{tikzcd}\ ] ] + ( in particular , @xmath47 can be factored into a reduction @xmath663 ( with image a complete section ) followed by a class - bijective homomorphism @xmath664 , or a surjective reduction @xmath241 followed by a class - injective homomorphism @xmath665 . ) 5 .
@xmath47 belongs to the smallest class of borel homomorphisms between countable borel equivalence relations which is closed under composition and contains all reductions and class - injective homomorphisms .
( i)@xmath453(ii ) is obvious .
( ii)@xmath453(iii ) : @xcite if @xmath658 is not smooth , then it has an ergodic invariant @xmath34-finite non - atomic measure @xmath147 .
the pushforward @xmath666 is then a @xmath205-ergodic ( because @xmath147 is @xmath667-ergodic ) measure on @xmath118 , hence concentrates at some @xmath435 , i.e. , @xmath668 , whence @xmath656 is not smooth .
( iii)@xmath453(iv ) : letting @xmath669 be the projection and @xmath240 be the equivalence relation on @xmath670 induced by @xmath1 , we have that @xmath659 is a surjective reduction , and @xmath47 descends along @xmath241 to a class - injective homomorphism @xmath671 . by [ thm : classinj - factor ] , @xmath672 factors as a complete section embedding @xmath660 followed by a class - bijective homomorphism @xmath661 , for some @xmath673 .
( iv)@xmath453(v ) is obvious .
( v)@xmath453(i ) : clearly reductions are smooth , as are class - bijective homomorphisms ; it follows that so are class - injective homomorphisms , by [ thm : classinj - factor ] ( see also @xcite ) . similarly
to before we now have [ thm : univstr - smh - red ] if @xmath69 is universally structurable , then @xmath674 , for all @xmath303 .
similarly , if @xmath39 is an elementary class , then @xmath675 .
[ thm : smh - elem ] let @xmath69 be a countable borel equivalence relation . then there is a sentence @xmath676 ( in some fixed language ) such that for all @xmath303 , we have @xmath677 iff @xmath678 . the language is @xmath679 where @xmath680 are unary predicates , and the sentence is @xmath681 , where @xmath652 is as in [ lm : sigmae ] , and @xmath682 it is easily seen that for any @xmath10-structure @xmath19 on @xmath43 , we will have @xmath683 iff the interpretation @xmath684 is a borel transversal of @xmath685 , where @xmath47 is the borel map to @xmath1 coded by @xmath19 . if @xmath448 is an elementary class ,
then so is @xmath686 .
this completes the proof of [ thm : red - elem ] .
[ thm : red - elem ] generalizes @xcite , which shows that some particular classes of the form @xmath74 , @xmath39 elementary , are closed under @xmath609 .
hjorth - kechris ( * ? ? ?
* d.3 ) proved that every @xmath74 ( @xmath39 elementary ) is closed under @xmath75 , i.e. , containment of equivalence relations on the same space . since containment is a class - injective homomorphism ( namely the identity ) , [ thm : red - elem ] also generalizes this .
see [ sec : fiber - factor ] for more on the relation between ( * ? ? ?
* appendix d ) and the above .
we end this section by pointing out that exactly analogous proofs work for yet another pair of ( `` global '' resp .
`` local '' ) classes of homomorphisms ( which we did not include in [ tbl : global - local ] ) , which forms a natural counterpart to @xmath612 .
we write @xmath687 to denote a ( borel ) class - surjective reduction , and @xmath688 to denote a class - surjective smooth homomorphism
. then we have [ thm : csr - elem ] let @xmath448 be an elementary class .
then @xmath689 are equal and elementary , and @xmath690 .
exactly as before , we have the following chain of results : [ thm : cssmh - factor ] let @xmath153 and @xmath691 .
then there is a @xmath615 , a surjective reduction @xmath659 , and a class - bijective homomorphism @xmath488 , such that @xmath617 . by [ thm : smh - factor ]
, @xmath47 can be factored into a surjective reduction @xmath241 followed by a class - injective homomorphism @xmath692 ; since @xmath693 is class - surjective and @xmath241 is surjective , @xmath692 must be class - surjective , i.e. , class - bijective .
[ thm : univstr - cssmh - csr ] if @xmath69 is universally structurable , then @xmath694 , for all @xmath303 .
similarly , if @xmath39 is an elementary class , then @xmath695 .
[ thm : cssmh - elem ] let @xmath69 .
then there is a sentence @xmath696 ( in some fixed language ) such that for all @xmath303 , we have @xmath697 iff @xmath698 . like [ thm : smh - elem ] , but put @xmath699 .
it follows that @xmath700 is elementary .
since any reduction @xmath701 can be factored into a surjective reduction onto its image followed by an embedding , we could have alternatively proved that @xmath74 is elementary ( for elementary @xmath39 ) by combining [ thm : emb - elem ] with [ thm : csr - elem ] .
we say that an elementary class @xmath66 is an elementary reducibility class if it is closed under reductions .
the following elementary classes mentioned in [ sec : elem - examples ] are elementary reducibility classes : smooth equivalence relations , hyperfinite equivalence relations , treeable equivalence relations @xcite , @xmath6 .
the following classes are _ not _ elementary reducibility classes : finite equivalence relations , aperiodic equivalence relations , compressible equivalence relations , compressible hyperfinite equivalence relations . in [ sec : freeact ] , we will prove that for a countably infinite group @xmath21 , @xmath702 is an elementary reducibility class iff @xmath21 is amenable , where @xmath702 consists of equivalence relations whose aperiodic part is generated by a free action of @xmath21 . by [ thm : red - elem ] , for every @xmath69 , @xmath703 is the smallest elementary reducibility class containing @xmath1 ; this is analogous to [ thm : elem - cone ] .
we also have the following analog of [ thm : elem - char ] : [ thm : elemred - char ] a class @xmath66 is an elementary reducibility class iff it is closed under smooth homomorphisms and contains an invariantly universal element @xmath67 , in which case @xmath704 . as well , there is the analog of [ thm : univstr ] : [ thm : sunivstr ] let @xmath69 .
the following are equivalent : 1 .
@xmath1 is invariantly universal in @xmath703 .
@xmath1 is invariantly universal in some elementary reducibility class .
3 . for every @xmath303 , @xmath678 iff @xmath185 .
we call @xmath69 stably universally structurable if these equivalent conditions hold .
we write @xmath705 for the class of stably universally structurable countable borel equivalence relations . for any @xmath69 , we write @xmath706 for the @xmath220-universal element of @xmath703 .
we emphasize that being stably universally structurable is a _ stronger _
notion than being universally structurable ( @xmath707 is a transversal of @xmath708 , which is a coarser equivalence relation than @xmath189 ) . in particular , note that `` stably universally structurable '' does _ not _ mean `` universal with respect to @xmath86 '' . by [ thm : smh - factor ]
, the preorder @xmath609 on @xmath6 is the composite @xmath709 of the two preorders @xmath86 and @xmath187 on @xmath6 , hence also the join of @xmath86 and @xmath187 in the complete lattice of all preorders on @xmath6 ( that are @xmath61-invariant , say ) , i.e. , @xmath609 is the finest preorder on @xmath6 coarser than both @xmath86 and @xmath187 .
similarly , @xmath708 is the join of @xmath188 and @xmath189 in the lattice of equivalence relations on @xmath6 ; this follows from noting that @xmath710 .
one may ask what is the meet of the preorders @xmath86 and @xmath187 .
we do not know of a simple answer .
note that the meet is strictly coarser than @xmath220 ; indeed , @xmath711 and @xmath712 , but @xmath713 .
( similarly , the meet of @xmath188 and @xmath189 is strictly coarser than @xmath61 . )
clearly one can define similar notions of `` elementary embeddability class '' and `` elementary class - surjective reducibility class '' .
dougherty - jackson - kechris proved several results relating borel reducibility to compressibility @xcite , which we state here in a form suited for our purposes .
[ thm : compress ] let @xmath226 be countable borel equivalence relations . 1
. @xmath1 is compressible iff @xmath714 ( and the latter is always compressible ) .
2 . if @xmath1 is compressible and @xmath172 , then @xmath184 .
3 . if @xmath43 is compressible and @xmath715 , then @xmath184 .
if @xmath226 are compressible and @xmath171 , then @xmath184 . 5 .
@xmath171 iff @xmath716 .
while these were all proved at some point in @xcite , not all of them were stated in this form . for ( a ) ,
see @xcite . for ( b ) , see @xcite .
clearly ( e ) follows from ( a ) and ( d ) ( and that @xmath717 ) .
we now sketch ( c ) and ( d ) , which are implicit in the proof of @xcite . for ( c ) , take @xmath718 , and let @xmath719 be the image of @xmath47 .
then @xmath47 is a surjective reduction @xmath720 , hence we can find a @xmath721 such that @xmath722 ; in particular , @xmath241 is a complete section embedding .
now @xmath240 is compressible , so applying @xcite , we get @xmath723 , whence @xmath724 . for ( d ) , take @xmath701 , and let @xmath725 be the image of @xmath47 . then @xmath726 and @xmath725 , whence @xmath727 . by ( a c ) , @xmath728 .
in passing , we note that [ thm : compress](b ) and [ thm : classinj - factor ] together give the following : if @xmath1 is compressible and @xmath729 , then @xmath181 .
it follows from [ thm : compress ] that the compressible equivalence relations ( up to isomorphism ) form a transversal of bireducibility , with corresponding selector @xmath730 , which is moreover compatible with the reducibility ordering .
we summarize this as follows .
let @xmath731 denote the compressible countable borel equivalence relations .
we have an order - isomorphism of posets @xmath732 [ thm : ein - not - clop - intop ] unlike the selector @xmath733 for @xmath189 , the selector @xmath730 for @xmath188 does not take @xmath1 to the @xmath220-greatest element of its @xmath188-class ( e.g. , @xmath734 ) . nor
does it always take @xmath1 to the @xmath220-least element of its @xmath188-class , or even to an element @xmath220-less than @xmath1 : for finite @xmath1 clearly @xmath735 , while for aperiodic @xmath1 , a result of thomas @xcite ( see also @xcite ) states that there are aperiodic @xmath1 such that @xmath736 .
we now relate compressibility to structurability .
let @xmath737 denote the invariantly universal compressible countable borel equivalence relation , i.e. , the @xmath220-universal element of @xmath738 . aside from @xmath730
, we have another canonical way of turning any @xmath1 into a compressible equivalence relation , namely @xmath739 .
these two maps are related as follows : [ thm : compress - univstr ] let @xmath69 be a countable borel equivalence relation .
1 . @xmath740 .
2 . suppose @xmath1 is universally structurable .
then : 1 .
@xmath741 is universally structurable , so @xmath742 ; 2 .
@xmath743 iff @xmath744 ( iff @xmath745 ) . for ( a ) , we have @xmath746 , whence @xmath747 .
now suppose @xmath1 is universally structurable .
then similarly to ( a ) , from @xmath748 we get @xmath742 , proving the latter part of ( i ) .
next we prove ( ii ) .
if @xmath744 , then @xmath748 gives @xmath749 .
conversely , if @xmath743 , then since @xmath741 is compressible , @xmath749 , and also @xmath750 , whence @xmath744 .
this proves ( ii ) ( the parenthetical follows from the latter part of ( i ) ) . finally , to finish proving ( i ) , find by [ thm : csr - elem ] a universally structurable @xmath751 such that @xmath752 is closed under class - surjective reductions .
then @xmath753 gives @xmath754 , whence by ( ii ) , @xmath755 is universally structurable .
[ rmk : aperiodic - univstr - compress ] we do not know if there is an aperiodic universally structurable @xmath1 with @xmath756 .
the example of thomas @xcite mentioned above is far from universally structurable , since it has a unique ergodic invariant probability measure .
we call a bireducibility class @xmath66 universally structurable if it contains a universally structurable element . in this case , by [ thm : red - elem ] , @xmath39 contains an invariantly universal ( stably universally structurable ) element , namely @xmath757 for any @xmath67 ; and by [ thm : compress - univstr ] , it also contains a compressible universally structurable element , namely @xmath741 for any @xmath67 . between these two ( in the ordering @xmath220 )
lie all those universally structurable @xmath67 such that @xmath743 .
let @xmath758 denote the class of compressible universally structurable equivalence relations .
since @xmath458 forms a transversal ( up to isomorphism ) of the equivalence relation @xmath189 , while @xmath738 forms a transversal of @xmath188 , we would expect @xmath759 to form a transversal of @xmath708 , the join of @xmath189 and @xmath188 .
that this is the case follows from the fact that the two corresponding selectors @xmath733 ( for @xmath760 ) and @xmath730 ( for @xmath188 ) commute : [ thm : einfty - e - in ] for any @xmath69 , @xmath761 .
we have @xmath762 , whence @xmath763 , and so @xmath764 .
conversely , we have @xmath765 , whence @xmath766 , and so @xmath767 , since the latter is universally structurable by [ thm : compress - univstr ] .
in this section , we consider the order - theoretic structure of the poset of elementary classes under inclusion ( equivalently the poset @xmath82 ) , as well as the poset of elementary reducibility classes under inclusion ( equivalently @xmath768 , or @xmath769 , or @xmath88 ) .
in [ sec : projection ] , we introduce some concepts from order theory which give us a convenient way of concisely stating several results from previous sections . in [ sec : lattice ] , we discuss meets and joins in the poset @xmath82 . in [ sec : pbr ] , we extend a well - known result of adams - kechris @xcite to show that @xmath88 is quite complicated , by embedding the poset of borel subsets of reals .
we remark that we always consider the empty equivalence relation @xmath770 on the empty set to be a countable borel equivalence relation ; this is particularly important in this section .
note that @xmath770 is ( vacuously ) @xmath34-structurable for any @xmath34 , hence is the @xmath220-universal @xmath771-structurable equivalence relation , where @xmath771 denotes an inconsistent theory . among the various posets ( or preordered sets ) of equivalence relations we have considered so far ( e.g. , @xmath248 , @xmath772 , @xmath773 ) ,
there is one which is both the finest and the most inclusive , namely @xmath774 .
several of the other posets and preorders may be viewed as derived from @xmath774 via the following general order - theoretic notion .
let @xmath775 be a poset .
a projection operator on @xmath776 is an idempotent order - preserving map @xmath777 , i.e. , @xmath778 the image @xmath779 of a projection operator @xmath780 is a retract of @xmath776 , i.e. , the inclusion @xmath781 has a one - sided ( order - preserving ) inverse @xmath782 , such that @xmath783 .
a projection operator @xmath780 also gives rise to an induced preorder @xmath784 on @xmath776 , namely the pullback of @xmath785 along @xmath780 , i.e. , @xmath786 leting @xmath787 , which is also the equivalence relation associated with @xmath784 , we thus have two posets derived from @xmath775 associated with each projection operator @xmath780 , namely the quotient poset @xmath788 and the subposet @xmath789 .
these are related by an order - isomorphism : @xmath790_\sim & |-- > e(x ) \\
e^{-1}(y ) = [ y]_\sim & < --| y.\end{aligned}\ ] ] ( there is the following analogy with equivalence relations : set @xmath791 poset , equivalence relation @xmath791 preorder , selector @xmath791 projection , and transversal @xmath791 retract . ) summarizing previous results , we list here several projection operators on @xmath774 that we have encountered , together with their images and induced preorders .
* @xmath733 , which has image @xmath792 ( the universally structurable equivalence relations ) and induces the preorder @xmath187 ( [ sec : univstr ] ) ; * @xmath730 , which has image @xmath793 ( the compressible equivalence relations ) and induces the preorder @xmath86 ( [ thm : compress ] ) ;
* @xmath794 ( [ thm : einfty - e - in ] ) , which has image @xmath795 ( the compressible universally structurable equivalence relations ) and induces the preorder @xmath609 ; * @xmath796 ( the @xmath220-universal element of @xmath703 ) , which has image @xmath797 ( the stably universally structurable equivalence relations ) and also induces the preorder @xmath609 ; * similarly , @xmath798 the @xmath220-universal element of @xmath799 , which induces @xmath606 . also note that some of these projection operators can be restricted to the images of others ; e.g. , the restriction of @xmath730 to @xmath458 is a projection operator on @xmath792 ( by [ thm : compress - univstr ] ) , with image @xmath795 .
again let @xmath775 be a poset , and let @xmath777 be a projection operator .
we say that @xmath780 is a closure operator if @xmath800 in other words , each @xmath801 is the ( @xmath785-)greatest element of its @xmath802-class . in that case , the induced preorder @xmath784 satisfies @xmath803 among the projection operators on @xmath774 listed above , three are closure operators , namely @xmath798 the @xmath220-universal element of @xmath451 , @xmath703 , or @xmath799 ( the first of these being @xmath733 ) . for another example
, let us say that a countable borel equivalence relation @xmath69 is idempotent if @xmath804 .
this is easily seen to be equivalent to @xmath805 ; hence , the idempotent elements of @xmath6 form the image of the closure operator @xmath806 on @xmath774 . note that all universally structurable equivalence relations are idempotent ( [ thm : univstr - idemp ] ) .
we now discuss the lattice structure of the poset of elementary classes under inclusion , equivalently the poset @xmath82 of universally structurable isomorphism classes under @xmath220 .
let us first introduce the following notation . for theories @xmath109 and @xmath807 , we write @xmath808 to mean that @xmath809 , i.e. , for every @xmath69 , if @xmath519 , then @xmath810 .
thus @xmath811 is a preorder on the class of theories which is equivalent to the poset of elementary classes ( via @xmath812 ) , and hence also to the poset @xmath82 ( via @xmath813 ) .
we denote the associated equivalence relation by @xmath814 . we stress that in the notation @xmath815 , @xmath34 and @xmath816 may belong to _ different _ languages
of course , if they happen to belong to the same language and @xmath34 logically implies @xmath816 , then also @xmath815 ; but the latter is in general a weaker condition .
let @xmath775 be a poset .
we say that @xmath776 is an @xmath83-complete lattice if every countable subset @xmath817 has a meet ( i.e. , greatest lower bound ) @xmath818 , as well as a join ( i.e. , least upper bound ) @xmath819 .
we say that @xmath776 is an @xmath83-distributive lattice if it is an @xmath83-complete lattice which satisfies the @xmath83-distributive laws @xmath820 where @xmath353 runs over a countable index set .
[ thm : lattice ] the poset @xmath82 is an @xmath83-distributive lattice , in which joins are given by @xmath821 , nonempty meets are given by @xmath822 , the greatest element is @xmath221 , and the least element is @xmath770 .
moreover , the inclusion @xmath84 preserves ( countable ) meets and joins . in other words , if @xmath823 are universally structurable equivalence relations , then @xmath596 ( respectively @xmath824 ) is their meet ( respectively join ) in @xmath774 as well as in @xmath82 . before giving the proof of [ thm : lattice ]
, we discuss the operations on theories which correspond to the operations @xmath822 and @xmath821 .
that is , let @xmath825 be a countable family of theories ; we want to find theories @xmath826 and @xmath827 such that @xmath828 and @xmath829 .
[ thm : tensor - thy ] let @xmath830 be the theory where @xmath831 is the disjoint union of the @xmath832 , and @xmath833 is the conjunction of the @xmath834 s regarded as being in the language @xmath831 ( so that the different @xmath834 s have disjoint languages )
. then @xmath835 . for each @xmath353 , the @xmath836-structure on @xmath837 has a reduct which is a @xmath834-structure , so @xmath838 , i.e. , @xmath839 ; hence @xmath840 .
conversely , for each @xmath351 we have @xmath841 so @xmath842 ; combining these @xmath843-structures yields a @xmath844-structure , so @xmath845 . while we can similarly prove that @xmath846 corresponds to the theory given by the disjunction of the @xmath834 s , we prefer to work with the following variant , which is slightly better behaved with respect to structurability .
let @xmath847 be the theory where @xmath848 where each is @xmath98 is a unary relation symbol , and @xmath849 ( where on the right - hand side , @xmath834 is regarded as having language @xmath850 ) . in other words
, @xmath851 asserts that for some ( unique ) @xmath353 , @xmath98 holds for all elements , and we have a @xmath834-structure ; and for all @xmath852 , @xmath853 and all relations in @xmath854 hold for no elements . then for a countable borel equivalence relation @xmath300 , a @xmath851-structure @xmath855 is the same thing as a borel @xmath1-invariant partition @xmath856 of @xmath2 , together with a @xmath834-structure @xmath857 for each @xmath353 .
[ thm : disjsum - thy ] @xmath858 .
the @xmath834-structure on each @xmath859 yields a @xmath860-structure ( with @xmath861 everything , and @xmath862 for @xmath852 ) ; so @xmath863 .
conversely , letting @xmath864 , we have @xmath865 and @xmath866 for each @xmath353 , whence @xmath867 . for a more formal viewpoint on the operations @xmath822 and @xmath821 on theories , see [ sec : interp - plus - times ] .
as noted in @xcite , the next lemma follows from abstract properties of the poset @xmath868 ; for the convenience of the reader , we include a direct proof .
[ thm : idemp - sum - join ] let @xmath869 be countably many idempotent countable borel equivalence relations
. then @xmath824 is their join in the preorder @xmath870 .
clearly each @xmath871 .
let @xmath303 and @xmath872 for each @xmath353 ; we must show that @xmath873 . since each @xmath872 , we have @xmath874 for some @xmath875 ; since @xmath876 , we have @xmath877 .
so an invariant embedding @xmath873 is built by invariantly embedding @xmath250 into @xmath878 so that the remainder ( complement of the image ) is isomorphic to @xmath43 , then similarly embedding @xmath879 into the remainder , etc . by [ thm : tensor - thy , thm : disjsum - thy ] , we may freely switch between the operations @xmath822 and @xmath821 on universally structurable equivalence relations , and the same operations on theories .
first we check that @xmath822 is meet and @xmath821 is join . let @xmath880 be theories ; it suffices to show that @xmath833 , resp .
, @xmath851 , is their meet , resp .
, join , in the preorder @xmath811 . for @xmath822
this is clear , since a @xmath881-structure on @xmath69 is the same thing as a @xmath834-structure for each @xmath353 . for @xmath821 , a @xmath834-structure on @xmath1 for any @xmath353 yields a @xmath882-structure ( corresponding to the partition of @xmath1 where the @xmath353th piece is everything ) ; thus @xmath883 for each @xmath353 . and
if @xmath807 is another theory with @xmath884 for each @xmath353 , then given a @xmath885-structure on @xmath1 , we have a partition of @xmath1 into pieces which are @xmath834-structured for each @xmath353 , so by @xmath884 we can @xmath816-structure each piece of the partition ; thus @xmath886 . that the inclusion @xmath887 preserves ( _ all _ existing ) meets follows from the fact that @xmath888 is the image of the closure operator @xmath733 .
that it preserves countable joins follows from [ thm : idemp - sum - join ] .
now we check the @xmath83-distributive laws .
distributivity of @xmath495 over @xmath821 follows from [ thm : tensor - props](e ) .
to check distributivity of @xmath889 over @xmath822 , we again work with theories .
let @xmath890 be theories ; we need to show that @xmath891 . the @xmath811 inequality , as in any lattice , is trivial .
for the converse inequality , let @xmath300 and @xmath892 , which amounts to a @xmath893 for each @xmath353 .
then for each @xmath353 , we have a borel @xmath1-invariant partition @xmath894 such that @xmath895 and @xmath896 . by combining the various @xmath897
, we get @xmath898 and @xmath899 ; and so the partition @xmath900 witnesses that @xmath901 .
it is not true that @xmath821 is join in @xmath774 , since there exist @xmath69 such that @xmath902 ( e.g. , @xmath903 ) .
similarly , it is not true that @xmath822 is meet in @xmath774 , since there are @xmath1 with @xmath904 ( see examples near end of [ sec : tensor ] ) . that the inclusion @xmath887 preserves countable joins suggests that perhaps @xmath888 is also the image of an `` interior operator '' .
this would mean that every countable borel equivalence relation @xmath69 contains ( in the sense of @xmath220 ) a greatest universally structurable equivalence relation .
we do not know if this is true . by restricting [ thm :
lattice ] to the class @xmath738 of compressible equivalence relations , which is downward - closed under @xmath220 , closed under @xmath821 , and has greatest element @xmath737 , we immediately obtain [ thm : lattice - compress ] the poset @xmath769 is an @xmath83-distributive lattice , in which joins are given by @xmath821 , nonempty meets are given by @xmath822 , the greatest element is @xmath737 , and the least element is @xmath770 .
moreover , the inclusion @xmath905 preserves ( countable ) meets and joins . now using that @xmath906 is isomorphic to @xmath85 , we may rephrase this as [ thm : lattice - red ] the poset of universally structurable bireducibility classes under @xmath86 is an @xmath83-distributive lattice .
moreover , the inclusion into the poset @xmath85 of all bireducibility classes under @xmath86 preserves ( countable ) meets and joins .
we stress that the @xmath86-meets in [ thm : lattice - red ] must be computed using the _ compressible _ elements of bireducibility classes .
that is , if @xmath226 are universally structurable , then their @xmath86-meet is @xmath907 , but _ not necessarily _ @xmath465 .
for example , if @xmath1 is invariantly universal finite and @xmath43 is invariantly universal aperiodic , then @xmath908 is clearly not the @xmath86-meet of @xmath226 . also , if there is an aperiodic universally structurable @xmath1 with @xmath756 , then ( by [ thm : compress - univstr ] ) @xmath909 is not the @xmath86-meet of @xmath1 and @xmath910 .
the order - theoretic structure of the poset @xmath85 of _ all _ bireducibility classes under @xmath86 is not well - understood , apart from that it is very complicated ( by @xcite ) .
the first study of this structure was made by kechris - macdonald in @xcite .
in particular , they raised the question of whether there exists any pair of @xmath86-incomparable @xmath87 for which a @xmath86-meet exists .
, together with the existence of many @xmath86-incomparable universally structurable bireducibility classes ( [ thm : pbr ] ) , answers this question by providing a large class of bireducibility classes for which @xmath86-meets always exist .
there are some natural order - theoretic questions one could ask about the posets @xmath911 and @xmath912 , which we do not know how to answer .
for example , is either a complete lattice ?
if so , is it completely distributive ?
is it a `` zero - dimensional '' @xmath83-complete lattice , in that it embeds ( preserving all countable meets and joins ) into @xmath913 for some set @xmath2 ? ( see [ thm : lattice-0d ] below for some partial results concerning this last question . )
it can be shown that every @xmath83-distributive lattice is a _ quotient _ of a sublattice of @xmath913 for some set @xmath2 ( see [ sec : sdlat ] ) .
in particular , this implies that the set of algebraic identities involving @xmath822 and @xmath821 which hold in @xmath911 is `` completely understood '' , in that it consists of exactly those identities which hold in @xmath914 .
we mention here some connections with recent work of marks @xcite .
let @xmath915 be countably many countable borel equivalence relations on the same standard borel space @xmath2 .
we say that the @xmath916 are independent if there is no sequence @xmath917 of distinct elements of @xmath2 , where @xmath918 , such that @xmath919 for some @xmath920 with @xmath921 for each @xmath351 . in that case , their independent join is the smallest equivalence relation on @xmath2 containing each @xmath916 .
for example , the independent join of treeable equivalence relations is still treeable .
marks proves the following for elementary classes closed under independent joins @xcite : [ thm : marks - ultrafilter ] if @xmath40 is an elementary class of aperiodic equivalence relations closed under binary independent joins , then for any borel homomorphism @xmath922 ( where @xmath2 is any standard borel space ) , there is some @xmath163 such that @xmath923 .
[ thm : marks - emb ] if @xmath40 is an elementary class of aperiodic equivalence relations closed under countable independent joins , then for any @xmath69 , if @xmath924 , then @xmath925 .
clearly the aperiodicity condition in [ thm : marks - ultrafilter , thm : marks - emb ] can be loosened to the condition that @xmath926 ( so that restricting @xmath40 to the aperiodic elements does not change @xmath62 up to biembeddability ) .
above we asked whether the @xmath83-distributive lattice @xmath88 is zero - dimensional , i.e. , embeds into @xmath913 for some set @xmath2 .
this is equivalent to asking whether there are enough @xmath83-prime filters ( i.e. , filters closed under countable meets whose complements are closed under countable joins ) in @xmath88 to separate points .
gives some examples of @xmath83-prime filters : [ thm : lattice-0d ] if @xmath40 contains @xmath236 and is closed under binary independent joins , then @xmath927 is an @xmath83-prime filter in @xmath88 .
if @xmath928 then we have a homomorphism @xmath929 sending the @xmath241-preimage of @xmath916 to @xmath353 ; by [ thm : marks - ultrafilter ] , it follows that @xmath930 for some @xmath353 .
we also have the following simple consequence of [ thm : marks - emb ] : if @xmath40 is an elementary class such that @xmath931 is closed under countable independent joins , then @xmath932 .
the @xmath220-universal element of @xmath931 reduces to @xmath62 , whence by [ thm : marks - emb ] it embeds into @xmath62 , i.e. , belongs to @xmath933 .
although the conclusions of [ thm : marks - ultrafilter , thm : marks - emb ] are invariant with respect to bireducibility ( respectively biembeddability ) , marks has pointed out that the notion of being closed under independent joins is not similarly invariant : there are @xmath934 such that @xmath40 is closed under independent joins but @xmath935 is not
. in particular , if @xmath34 axiomatizes trees while @xmath816 axiomatizes trees of degree @xmath936 , then @xmath937 by @xcite ; but it is easy to see ( using an argument like that in [ thm : indepjoin - treeable ] below ) that independent joins of @xmath816-structurable equivalence relations can have arbitrarily high cost , so are not all @xmath816-structurable .
clearly if @xmath938 are closed under independent joins , then so is @xmath939 .
in particular , the class @xmath738 of compressible equivalence relations is closed under independent joins , since the join of compressible equivalence relations contains a compressible equivalence relation ; thus the class of compressible treeable equivalence relations is closed under independent joins .
we note that this is the smallest nontrivial elementary class to which [ thm : marks - ultrafilter , thm : marks - emb ] apply : [ thm : indepjoin - treeable ] if @xmath40 is an elementary class containing @xmath236 and closed under binary independent joins , then @xmath40 contains all compressible treeable equivalence relations .
since @xmath40 is elementary and contains @xmath236 , it contains all aperiodic smooth countable borel equivalence relations .
now let @xmath300 be compressible treeable .
by @xcite , there is a borel treeing @xmath940 with degree @xmath936 .
by @xcite ( see also remarks following @xcite ) , there is a borel edge coloring @xmath941 .
then @xmath1 is the independent join of the equivalence relations @xmath942 for @xmath943 .
since the @xmath916 are not aperiodic , consider the following modification .
let @xmath944 let @xmath945 be the tree on @xmath946 consisting of @xmath329 on @xmath2 and the edges @xmath947 and @xmath948 for @xmath163 , @xmath949 , and @xmath253 , and let @xmath950 extend @xmath951 with @xmath952 ( note that @xmath953 is not an edge coloring ) .
then the inclusion @xmath954 is a complete section embedding of each @xmath916 into the equivalence relation @xmath955 generated by @xmath956 , and of @xmath1 into the equivalence relation @xmath957 generated by @xmath945 .
it follows that each @xmath955 is ( aperiodic ) smooth ( because @xmath916 is ) , hence in @xmath40 , while @xmath958 ( because @xmath1 is compressible ) .
but it is easily seen that @xmath957 is the independent join of the @xmath955 , whence @xmath299 .
adams - kechris @xcite proved the following result showing that the poset @xmath85 is extremely complicated : there is an order - embedding from the poset of borel subsets of @xmath89 under inclusion into the poset @xmath85 . in this short section , we show that their proof may be strengthened to yield [ thm : pbr ] there is an order - embedding from the poset of borel subsets of @xmath89 under inclusion into the poset @xmath88 .
by @xcite , there is a countable borel equivalence relation @xmath133 , a borel homomorphism @xmath959 , and a borel map @xmath960 taking each @xmath961 to a borel probability measure @xmath962 on @xmath2 , such that , putting @xmath963 , we have 1 . for each @xmath961 , @xmath962 is nonatomic , concentrated on @xmath964 , @xmath965-invariant , and @xmath965-ergodic ; 2 .
if @xmath966 with @xmath967 , then every borel homomorphism @xmath968 maps a borel @xmath965-invariant set @xmath969 of @xmath962-measure @xmath369 to a single @xmath970-class . for borel @xmath971 , put @xmath972 and @xmath973 .
we claim that @xmath974 gives the desired order - embedding .
it is clearly order - preserving .
now suppose @xmath975 with @xmath976 but @xmath977 . by taking @xmath978
, we get @xmath979 but @xmath980 . let @xmath981 , and let @xmath982 be the second projection
. then @xmath983 , whence by @xmath965-ergodicity of @xmath962 , there is a @xmath984 and an @xmath965-invariant @xmath969 of @xmath962-measure @xmath369 such that @xmath985 , i.e. , @xmath986 . by ( ii ) above , there is a further @xmath965-invariant @xmath987 of @xmath962-measure @xmath369 such that @xmath988 is contained in a single @xmath970-class .
but since @xmath989 is class - bijective and @xmath47 is a reduction , this implies that @xmath990 is smooth , a contradiction .
if in [ thm : pbr ] we replace @xmath88 with @xmath82 ( thus weakening the result ) , then a simpler proof may be given , using groups of different costs ( see @xcite ) instead of @xcite .
the picture below is a simple visualization of the poset @xmath774 . for the sake of clarity , among the hyperfinite equivalence relations ,
only the aperiodic ones are shown .
six landmark universally structurable equivalence relations are shown ( circled dots ) : @xmath770 , @xmath991 ( @xmath220-universal aperiodic smooth ) , @xmath213 ( @xmath220-universal compressible hyperfinite ) , @xmath992 ( @xmath220-universal aperiodic hyperfinite ) , @xmath737 ( @xmath220-universal compressible ) , and @xmath221 ( @xmath220-universal ) .
also shown is the `` backbone '' of compressible equivalence relations ( bold line ) , which contains one element from each bireducibility class ( dotted loops ) .
the middle of the picture shows a `` generic '' universally structurable @xmath1 and its relations to some canonical elements of its bireducibility class : the @xmath220-universal element @xmath757 and the compressible element @xmath741 .
note that @xmath741 is not depicted as being below @xmath1 , in accordance with [ rmk : aperiodic - univstr - compress ] .
note also that for non - smooth @xmath1 , the @xmath220-universal element @xmath757 of its bireducibility class would indeed be above @xmath992 , as shown : @xmath993 implies @xmath994 since @xmath757 is stably universally structurable . finally , note that the picture is somewhat misleading in a few ways .
it is not intended to suggest that the compressibles form a linear order . nor is it intended that any of the pairs @xmath995 do not have anything strictly in between them ( except of course for the things below @xmath992 , which are exactly as shown ) .
let @xmath21 be a countably infinite group .
recall ( from [ sec : elem - examples ] ) that we regard @xmath21 as a structure in the language @xmath285 , where @xmath996 is the graph of the left multiplication of @xmath997 on @xmath21 .
thus , @xmath998 ( where @xmath288 is the scott sentence of @xmath21 in @xmath999 ) is the class of borel equivalence relations generated by a free borel action of @xmath21 . our main goal in this section is to characterize when @xmath76 is an elementary reducibility class . actually , to deal with a technicality , we need to consider the following variant of @xmath76 .
let @xmath1000 , where @xmath1001 is a sentence axiomatizing the finite equivalence relations .
thus @xmath702 consists of countable borel equivalence relations whose aperiodic part is generated by a free borel action of @xmath21 .
this is needed because every equivalence relation in @xmath76 must have all classes of the same cardinality as @xmath21 .
[ thm : freeact ] let @xmath21 be a countably infinite group .
the following are equivalent : 1 .
@xmath21 is amenable .
@xmath702 is closed under @xmath186 .
@xmath702 is closed under @xmath86 , i.e. , @xmath702 is an elementary reducibility class . to motivate [ thm : freeact ] , consider the following examples . by [ thm : hyperfinite ]
, @xmath1002 is the class of all hyperfinite equivalence relations , which is closed under @xmath86 . on the other hand , for every @xmath1003
, the free group @xmath1004 on @xmath104 generators is such that @xmath1005 is the class of treeable equivalence relations , by @xcite ; but @xmath1006 is not itself the class of all treeables , since every @xmath1007 with a nonatomic invariant probability measure has cost @xmath104 ( see @xcite ) .
recall that the @xmath220-universal element of @xmath76 is @xmath384 , the orbit equivalence of the free part of the shift action of @xmath21 on @xmath77 .
thus the @xmath220-universal element of @xmath702 is @xmath1008 , where @xmath1009 is the @xmath220-universal finite equivalence relation ( given by @xmath1010 ) .
seward and tucker - drob @xcite have shown that for countably infinite @xmath21 , every free borel action of @xmath21 admits an equivariant class - bijective map into @xmath1011 ( clearly the same holds for finite @xmath21 ) .
it follows that @xmath1011 is @xmath187-universal in @xmath76 .
a well - known open problem asks whether every orbit equivalence of a borel action of a countable amenable group @xmath21 is hyperfinite . in the purely borel context ,
the best known general result is the following @xcite : [ thm : locally - nilpotent - hyperfinite ] if @xmath21 is a countable locally nilpotent group , i.e. , every finitely generated subgroup of @xmath21 is nilpotent , then every orbit equivalence @xmath136 of a borel action of @xmath21 is hyperfinite .
recently conley , jackson , marks , seward , and tucker - drob have found examples of solvable but not locally nilpotent countable groups for which the conclusion of [ thm : locally - nilpotent - hyperfinite ] still holds .
if [ thm : locally - nilpotent - hyperfinite ] generalizes to arbitrary countable amenable @xmath21 , then it would follow that @xmath702 is the class of all hyperfinite equivalence relations ( since it contains @xmath384 which admits an invariant probability measure ) ; then the main implication ( i)@xmath453(iii ) in [ thm : freeact ] would trivialize . in the measure - theoretic context , a classical result of ornstein - weiss @xcite states that the orbit equivalence of a borel action of an amenable group @xmath21 is hyperfinite almost everywhere with respect to every probability measure .
we will need a version of this result which is uniform in the measure , which we now state . for a standard borel space @xmath2 ,
we let @xmath1012 denote the space of probability borel measures on @xmath2 ( see ( * ? ? ?
* e ) ) .
[ thm : ow - uniform ] let @xmath193 be standard borel spaces , @xmath1013 be the orbit equivalence of a borel action of a countable amenable group @xmath21 on @xmath2 , and @xmath1014 . then there is a borel set @xmath1015 , with @xmath1016 ( where @xmath1017 is the first projection ) , such that 1 . for each @xmath435 , @xmath1018 has @xmath1019-measure @xmath369 and is @xmath1-invariant ; 2 .
@xmath1020 is hyperfinite .
this follows from verifying that the proofs of @xcite can be made uniform .
we omit the details , which are tedious but straightforward .
we now have the following , which forms the core of [ thm : freeact ] : [ thm : amenact - disjsum ] let @xmath21 be a countable amenable group , and let @xmath153 be countable borel equivalence relations . if @xmath171 and @xmath1021 for some borel action of @xmath21 on @xmath118 , then @xmath1 is the disjoint sum of a hyperfinite equivalence relation and a compressible equivalence relation .
if @xmath1 is compressible , then we are done .
otherwise , @xmath1 has an invariant probability measure .
consider the ergodic decomposition of @xmath1 ; see e.g. , @xcite .
this gives a borel homomorphism @xmath1022 such that 1 .
@xmath1023 is a surjection onto the borel set @xmath1024 of ergodic invariant probability measures on @xmath1 ; 2 . for each @xmath1025
, we have @xmath1026 .
let @xmath701 , and apply [ thm : ow - uniform ] to @xmath43 and @xmath1027 , where @xmath1028 is the pushforward of measures .
this gives borel @xmath1029 such that 1 .
for each @xmath1025 , @xmath1030 , and @xmath1031 is @xmath43-invariant ( so @xmath27 is @xmath1032-invariant ) ; 2 .
@xmath1033 is hyperfinite .
now consider the homomorphism @xmath1034 , i.e. , @xmath1035 .
then @xmath241 is a reduction because @xmath47 is .
it follows that @xmath1036 is @xmath1-invariant and @xmath1037 is hyperfinite .
it now suffices to note that @xmath1038 is compressible .
indeed , otherwise it would have an ergodic invariant probability measure , i.e. , there would be some @xmath1025 such that @xmath1039 .
but then @xmath1040 , while @xmath1041 , a contradiction . clearly ( iii)@xmath453(ii ) .
if ( ii ) holds , then @xmath1042 implies that @xmath250 is generated by a free action of @xmath21 , so since @xmath250 is hyperfinite and has an invariant probability measure , @xmath21 is amenable ( see @xcite ) ) .
so it remains to prove ( i)@xmath453(iii ) .
let @xmath1043 .
then @xmath1 splits into a smooth part , which is clearly in @xmath702 , and a part which reduces to some @xmath1044 ; so we may assume @xmath1045 .
factor the reduction @xmath171 into a surjective reduction @xmath1046 ( onto the image ) followed by an embedding @xmath725 . by [ thm : amenact - disjsum ] , @xmath1047 , where @xmath1048 is hyperfinite and @xmath1049 is compressible
. then @xmath1050 .
since @xmath1051 and @xmath1049 is compressible , we have @xmath1052 ( [ thm : compress ] ) and so @xmath1053
. finally , we have @xmath1054 , since @xmath702 contains all hyperfinite equivalence relations ( because @xmath1055 , by ornstein - weiss s theorem and [ thm : hyperfinite ] ) .
in the previous sections , we have studied the relationship between structurability and common notions from the theory of countable borel equivalence relations .
this section , by contrast , concerns the other side of the @xmath1056 relation , i.e. , logic .
in particular , we are interested in model - theoretic properties of theories @xmath109 which are reflected in the elementary class @xmath40 that they axiomatize . a general question one could ask is when two theories @xmath1057 axiomatize the same elementary class , i.e. , in the notation of [ sec : lattice ] , when does @xmath1058 .
our main result here answers one instance of this question .
let @xmath1059 denote any sentence axiomatizing the smooth countable borel equivalence relations .
[ thm : einftys - smooth ] let @xmath109 be a theory .
the following are equivalent : 1 .
there is an @xmath33-formula @xmath95 which defines a finite nonempty subset in any countable model of @xmath34 .
2 . @xmath1060 , i.e. , any @xmath34-structurable equivalence relation is smooth , or equivalently @xmath62 is smooth .
3 . for any countably infinite group @xmath21
, we have @xmath1061 , i.e. , any @xmath34-structurable equivalence relation generated by a free borel action of @xmath21 is smooth .
4 . there is a countably infinite group @xmath21 such that @xmath1061 .
in particular , this answers a question of marks ( * ? ? ? * end of section 4.3 ) , who asked for a characterization of when @xmath94 ( @xmath93 a scott sentence ) is smooth .
the proof uses ideas from topological dynamics and ergodic theory .
marks observed that recent work of ackerman - freer - patel @xcite implies the following sufficient condition for a structure @xmath19 to structure every aperiodic countable borel equivalence relation . in [ sec : einftys - univ ] , we present his proof of this result , as well as several corollaries and related results .
the result refers to the model - theoretic notion of trivial definable closure ; see [ sec : einftys - univ ] for details .
let @xmath1062 denote any sentence axiomatizing the aperiodic countable borel equivalence relations .
[ thm : einftya - univ ] let @xmath10 be a language and @xmath19 be a countable @xmath10-structure with trivial definable closure .
then @xmath1063 , i.e. , every aperiodic countable borel equivalence relation is @xmath19-structurable .
in [ sec : scott ] we discuss the problem of when an elementary class can be axiomatized by a scott sentence .
we now begin the proof of [ thm : einftys - smooth ] .
the implication ( i)@xmath453(ii ) is easy : given a formula @xmath1064 as in ( i ) , @xmath1064 may be used to uniformly pick out a finite nonempty subset of each @xmath62-class , thus @xmath62 is smooth .
the implications ( ii)@xmath453(iii)@xmath453(iv ) are obvious .
so let @xmath21 be as in ( iv ) .
consider the logic action of @xmath1065 on @xmath1066 , the space of @xmath10-structures with universe @xmath21 . recall that this is given as follows : for @xmath1067 , @xmath1068 , @xmath104-ary @xmath111 , and @xmath1069 , we have @xmath1070 we regard @xmath21 as a subgroup of @xmath1065 via the left multiplication action , so that @xmath21 acts on @xmath1066 .
[ lm : modgamma - smooth ] let @xmath21 be a countably infinite group .
then @xmath1061 iff @xmath1071 is smooth .
the proof is essentially that of @xcite , with an additional wrinkle in the use of @xcite below ( to deal with @xmath34-finite measures ) .
@xmath1072 : suppose @xmath133 is generated by a free borel action of @xmath21 , @xmath1073 , and @xmath1 is not smooth ( hence witnessing that @xmath1074 ) .
then it admits an ergodic non - atomic invariant ( @xmath34-finite ) measure @xmath147 .
define @xmath1075 by @xmath1076 then @xmath47 is @xmath21-equivariant , whence @xmath666 is an ergodic non - atomic invariant ( @xmath34-finite ) measure on @xmath1071 , which is therefore not smooth .
@xmath1077 : let @xmath1078 be an ergodic non - atomic invariant measure on @xmath1071 .
consider the free part @xmath1079 of the shift action of @xmath21 on @xmath1080 , with orbit equivalence @xmath1081 .
the usual product measure @xmath1082 on @xmath1080 concentrates on @xmath118 , and is invariant and mixing with respect to the action of @xmath21 on @xmath118 ( see @xcite ) .
then consider the product action of @xmath21 on @xmath1083 , which is free since @xmath21 acts freely on @xmath118 .
by @xcite , this product action admits @xmath1084 as an ergodic non - atomic invariant ( @xmath34-finite ) measure . thus @xmath1085 is not smooth .
now observe that @xmath1085 is the skew product @xmath1086 with respect to the cocycle
@xmath1087 associated to the free action of @xmath21 on @xmath118 ; and that @xmath275 , when regarded as a cocycle @xmath1088 , is induced , in the sense of [ rmk : skew - product ] , by @xmath1089 where @xmath1090 .
so ( as in the proof of [ thm : esigma ] ) @xmath1085 is @xmath34-structurable , hence witnesses that @xmath1074 .
so we have converted ( iv ) in [ thm : einftys - smooth ] into a property of the action of @xmath21 on @xmath1091 .
our next step requires some preparation .
let @xmath10 be a language and @xmath110 be a countable @xmath10-structure .
we say that @xmath19 has the weak duplication property ( wdp ) if for any finite sublanguage @xmath1092 and finite subset @xmath1093 , there is a finite subset @xmath1094 disjoint from @xmath43 such that @xmath1095 ( here @xmath1096 denotes the reduct in the sublanguage @xmath1097 ) .
if we define the duplication property ( dp ) for @xmath19 by replacing @xmath1097 in the above by @xmath10 , then clearly the dp is equivalent to the strong joint embedding property ( sjep ) for the age of @xmath19 : for any @xmath1098 , there is @xmath1099 and embeddings @xmath1100 and @xmath1101 with disjoint images .
( recall that @xmath1102 is the class of finite @xmath10-structures embeddable in @xmath19 . ) for a countable group @xmath21 acting continuously on a topological space @xmath2 , we say that a point @xmath163 is recurrent if @xmath164 is not isolated in the orbit @xmath1103 ( with the subspace topology ) .
when @xmath2 is a polish space , a basic fact is that @xmath136 is smooth iff it does not have a recurrent point ; see e.g. , @xcite . thus far , we have only regarded the space @xmath29 of @xmath10-structures on a countable set @xmath2 as a standard borel space . below we will also need to consider the topological structure on @xmath29 .
see e.g. , @xcite .
in particular , we will use the system of basic clopen sets consisting of @xmath1104 where @xmath1092 is a finite sublanguage and @xmath1105 is an @xmath1097-structure on a finite nonempty subset @xmath1093 .
the next lemma , which translates between the dynamics of @xmath1091 and a model - theoretic property of @xmath34 , is the heart of the proof of ( iv)@xmath453(i ) in [ thm : einftys - smooth ] : [ lm : modgamma - recurrent ] let @xmath21 be a countably infinite group .
let @xmath10 be a language , and let @xmath34 be an @xmath33-sentence such that @xmath1106 is a @xmath1107 subspace .
suppose there is a countable model @xmath1108 with the wdp , such that the interpretation @xmath1109 of some @xmath1110 is not definable ( without parameters ) from equality .
then the action of @xmath21 on @xmath1091 has a recurrent point , thus @xmath1071 is not smooth .
we claim that it suffices to show that 1 .
every basic clopen set @xmath1111 containing some isomorphic copy of @xmath19 also contains two distinct isomorphic copies , @xmath1112 and @xmath1113 , from the same @xmath21-orbit .
suppose this has been shown ; we complete the proof .
note that since @xmath19 has wdp , @xmath19 must be infinite .
let @xmath1114 denote the closure in @xmath1091 of @xmath1115 ( where @xmath93 is the scott sentence of @xmath19 ) .
since @xmath1106 is @xmath1107 , @xmath1114 is a polish space , which is nonempty because it contains an isomorphic copy of @xmath19 . for each basic clopen set @xmath1111 , the set of @xmath1116 such that @xmath1117 is clearly open ; and by ( @xmath398 ) , it is also dense
. thus the set of recurrent points in @xmath1114 , i.e. , the set of @xmath1116 for which ( @xmath399 ) holds for _ every _
@xmath1118 , is comeager .
so it remains to prove ( @xmath398 ) .
let @xmath1119 be such that @xmath1118 contains an isomorphic copy of @xmath19 .
let @xmath110 , and let @xmath1105 where @xmath1120 is finite nonempty and @xmath1092 is finite .
we may assume @xmath1121 .
since @xmath1118 contains an isomorphic copy of @xmath19 , there is a map @xmath1122 which is an embedding @xmath1123 .
we will extend @xmath47 to a bijection @xmath1124 , and then define @xmath1125 , thus ensuring that @xmath1126 ; we need to choose @xmath47 appropriately so that there is @xmath139 with @xmath1127 .
put @xmath1128 . by wdp , there is a @xmath1129 disjoint from @xmath240 such that @xmath1130 , say via @xmath1131 . by the hypothesis that @xmath1109 is not definable from equality , there are @xmath1132 and @xmath1133 , where @xmath104 is the arity of @xmath1134 , such that @xmath1135 , @xmath1136 , and @xmath1137 have the same equality type , i.e. , we have a bijection @xmath1138 sending @xmath1139 to @xmath1140 . again by wdp
, we may find @xmath1137 disjoint from @xmath240 , @xmath1048 , and each other .
now pick @xmath1141 disjoint from @xmath43 and with the same equality type as @xmath1142 , and pick @xmath139 such that @xmath1143 and @xmath1144 are disjoint from @xmath43 and @xmath1145 .
extend @xmath1122 to a bijection @xmath1146 such that @xmath1147 then putting @xmath1125 , it is easily verified that @xmath1148 , i.e. , @xmath1149 ; but @xmath1150 , so @xmath1113 .
[ lm : einftys - smooth - wdp ] let @xmath21 be a countably infinite group .
let @xmath10 be a language , and let @xmath34 be an @xmath33-sentence such that @xmath1106 is a @xmath1107 subspace . if @xmath1061 , then no countable model of @xmath34 has the wdp .
suppose a countable ( infinite ) @xmath1108 has the wdp .
if for some @xmath1110 , the interpretation @xmath1109 is not definable ( without parameters ) from equality , then @xmath1074 by [ lm : modgamma - smooth ] and [ lm : modgamma - recurrent ] .
otherwise , clearly any aperiodic countable borel equivalence relation is @xmath93-structurable , hence @xmath34-structurable ; taking @xmath1011 then yields that @xmath1074 . working towards ( i ) in [ thm : einftys - smooth ] , which asserts the existence of a formula with certain properties , we now encode the wdp into formulas , using the following combinatorial notion .
let @xmath2 be a set and @xmath1151 .
an @xmath104-ary intersecting family on @xmath2 is a nonempty collection @xmath1152 of subsets of @xmath2 of size @xmath104 such that every pair @xmath1153 has @xmath1154 .
[ lm : wdp - nif ] let @xmath10 be a language .
there are @xmath33-formulas @xmath1155 for each @xmath1151 , such that for any countable @xmath10-structure @xmath110 without the wdp , there is some @xmath104 such that @xmath1156 is an @xmath104-ary intersecting family on @xmath2 .
let @xmath1157 enumerate all countably many triples where @xmath1158 is a finite sublanguage , @xmath1159 , and @xmath1160 is an @xmath1161-structure with universe @xmath1162 (= @xmath1163 ) . for each @xmath664 ,
let @xmath1164 be an @xmath1161-formula asserting that @xmath1165 ( are pairwise distinct and ) form an @xmath1161-substructure isomorphic to @xmath1166 , which is not disjoint from any other such substructure .
thus @xmath19 does not have the wdp iff some @xmath1167 holds for some tuple in @xmath19 ; and in that case , the collection of all tuples ( regarded as sets ) for which @xmath1167 holds will form an @xmath1162-ary intersecting family . finally put @xmath1168 so that @xmath1169 is equivalent to @xmath1167 for the least @xmath664 which holds for some tuple . recall that ( i ) in [ thm : einftys - smooth ] asserts the existence of a single formula defining a finite nonempty set .
the following lemma , due to clemens - conley - miller @xcite , gives a way of uniformly defining a finite nonempty set from an intersecting family .
for the convenience of the reader , we include its proof here .
[ thm : nif - fin ] let @xmath1152 be an @xmath104-ary intersecting family on @xmath2 . for @xmath1170
, define @xmath1171 then there exist @xmath1172 such that @xmath1173 are ( respectively @xmath1174-ary , @xmath1175-ary , etc . ) intersecting families , and @xmath1176 is finite .
it suffices to show that if @xmath1152 is infinite , then there is some @xmath1170 such that @xmath1177 is an @xmath262-ary intersecting family .
indeed , having shown this , we may find the desired @xmath1178 inductively ; the process must terminate since a @xmath369-ary intersecting family is necessarily a singleton .
so assume @xmath1152 is infinite , and let @xmath1179 be greatest so that @xmath1177 is nonempty . let @xmath1180 . for each @xmath1181 , by our choice of @xmath262 , there are only finitely many @xmath1182 such that @xmath1183 .
thus by definition of @xmath1177 , there is @xmath1182 such that @xmath1184 and @xmath1185 . similarly , there is @xmath1186 such that @xmath1187 and @xmath1188
. then @xmath1189 , as desired .
[ lm : wdp - fin ] let @xmath10 be a language .
there is an @xmath33-formula @xmath95 such that for any countable @xmath10-structure @xmath19 without the wdp , @xmath1190 is a finite nonempty subset .
this follows from [ lm : wdp - nif ] and [ thm : nif - fin ] , by a straightforward encoding of the operation @xmath1191 in @xmath33 . in more detail
, for each @xmath33-formula @xmath1192 and @xmath1179 , let @xmath1193 be a formula asserting that @xmath1194 are pairwise distinct and there are infinitely many extensions @xmath1195 such that @xmath1192 holds , so that if @xmath1196 defines ( in the sense of [ lm : wdp - nif ] ) a family @xmath1152 of subsets of size @xmath104 , then @xmath1197 defines @xmath1177 .
let @xmath1198 for @xmath1151 be given by [ lm : wdp - nif ] .
for each finite tuple @xmath1199 such that @xmath1200 , let @xmath1201 be a sentence asserting that @xmath1202 holds for at least one but only finitely many tuples . then letting @xmath1203 enumerate all such tuples , the desired formula @xmath1064 can be given by @xmath1204 by [ lm : wdp - nif , thm : nif - fin ] , in any countable @xmath10-structure @xmath19 without the wdp , @xmath1190 will be the union of the finitely many sets in some intersecting family .
[ lm : einftys - smooth - gdelta ] let @xmath21 be a countably infinite group .
let @xmath10 be a language , and let @xmath34 be an @xmath33-sentence such that @xmath1106 is a @xmath1107 subspace .
if @xmath1061 , then there is an @xmath33-formula @xmath95 which defines a finite nonempty subset in any countable model of @xmath34 . by [ lm : einftys - smooth - wdp ] and [ lm : wdp - fin ] .
to complete the proof of [ thm : einftys - smooth ] , we need to remove the assumption that @xmath1106 is @xmath1107 from [ lm : einftys - smooth - gdelta ] .
this can be done using the standard trick of morleyization , as described for example in ( * ? ? ?
* section 2.6 ) for finitary first - order logic , or @xcite for @xmath33 . given any language @xmath10 and @xmath33-sentence @xmath34 , by adding relation symbols for each formula in a countable fragment of @xmath33 containing the sentence @xmath34
, we obtain a new ( countable ) language @xmath1097 and an @xmath1205-sentence @xmath1206 such that * the @xmath10-reduct of every countable model of @xmath1206 is a model of @xmath34 ; * every countable model of @xmath34 has a unique expansion to a model of @xmath1206 ; * @xmath1206 is ( logically equivalent to a formula ) of the form @xmath1207 where each @xmath1208 is a quantifier - free finitary @xmath1097-formula , whence @xmath1209 is @xmath1107 .
it follows that the conditions ( i ) and ( iv ) in [ thm : einftys - smooth ] for @xmath109 are equivalent to the same conditions for @xmath826 .
so [ lm : einftys - smooth - gdelta ] holds also without the assumption that @xmath1106 is @xmath1107 , which completes the proof of [ thm : einftys - smooth ] .
we conclude this section by pointing out the following analog of [ lm : modgamma - smooth ] : [ lm : modgamma - compress ] let @xmath21 be a countably infinite group and @xmath109 be a theory .
then @xmath1210 iff @xmath1071 is compressible .
exactly as in [ lm : modgamma - smooth ] , but using probability measures instead of @xmath34-finite measures .
this has the following corollaries .
the first strengthens ( * ? ? ?
* section 6.1.10 ) : let @xmath279 denote the class of trees , and more generally , let @xmath280 denote the class of contractible @xmath104-dimensional simplicial complexes .
then for each @xmath104 , there is some countably infinite group @xmath21 such that @xmath1211 admits no @xmath21-invariant measure ( and thus no @xmath1065-invariant measure ) . for each @xmath104 ,
let @xmath1212 be a sentence axiomatizing @xmath280 . for @xmath1213 , take @xmath21 to be any infinite kazhdan group .
by @xcite , no free borel action of @xmath21 admitting an invariant probability measure is treeable , i.e. , @xmath1214 ; thus @xmath1211 admits no @xmath21-invariant measure by [ lm : modgamma - compress ] . for @xmath1215 ,
take @xmath1216 . by a result of gaboriau ( see , e.g. , @xcite ) , no free borel action of @xmath21 admitting an invariant probability measure can be @xmath280-structurable .
let @xmath10 be a language , and let @xmath34 be an @xmath33-sentence such that @xmath1217 is a closed subspace .
then for any countably infinite group @xmath21 , there is a free borel action of @xmath21 which admits an invariant probability measure and is @xmath34-structurable .
since @xmath1106 is closed , it is compact , so since @xmath1065 is amenable , @xmath1091 admits a @xmath1065-invariant probability measure , thus a @xmath21-invariant probability measure ; then apply [ lm : modgamma - compress ] .
several theories @xmath109 are known to axiomatize @xmath6 , the class of all countable borel equivalence relations .
for example , by ( * ? ? ?
* ( proof of ) 3.12 ) , every @xmath69 is structurable via locally finite graphs .
more generally , one can consider @xmath34 such that every aperiodic or compressible countable borel equivalence relation is @xmath34-structurable .
for example , it is folklore that every aperiodic countable borel equivalence relation can be structured via dense linear orders ( this will also follow from [ thm : einftya - univ ] below ) , while the proof of @xcite shows that every compressible @xmath69 is structurable via graphs of vertex degree @xmath936 .
a result that some particular @xmath34 axiomatizes @xmath6 ( or all aperiodic @xmath1 ) shows that every ( aperiodic ) @xmath69 carries a certain type of structure , which can be useful in applications .
a typical example is the very useful marker lemma ( see @xcite ) , which shows that every aperiodic @xmath1 admits a decreasing sequence of borel complete sections @xmath96 with empty intersection .
this can be phrased as : every aperiodic countable borel equivalence relation @xmath1 is @xmath34-structurable , where @xmath34 in the language @xmath97 asserts that each ( unary ) @xmath98 defines a nonempty subset , @xmath99 , and @xmath100 .
we now give the proof of [ thm : einftya - univ ] , which provides a large class of examples of such theories .
to do so , we first review the main result from @xcite .
let @xmath10 be a language and @xmath110 be a countable @xmath10-structure .
we say that @xmath19 has trivial definable closure ( tdc ) if the following equivalent conditions hold ( see @xcite , @xcite ) : * for every finite @xmath1093 , the pointwise stabilizer @xmath1218 of the automorphism group of @xmath19 on @xmath43 fixes no other elements of @xmath2 ; * for every finite @xmath1093 , the pointwise stabilizer @xmath1218 of the automorphism group of @xmath19 on @xmath43 has infinite orbits on @xmath1219 ( trivial algebraic closure ) ; * for every finite @xmath1093 and @xmath33-formula @xmath95 with parameters in @xmath43 , if there is a unique @xmath163 such that @xmath1220 holds , then @xmath1221 ; * for every finite @xmath1093 and @xmath33-formula @xmath95 with parameters in @xmath43 , if there are only finitely many @xmath1222 such that @xmath1223 holds , then @xmath1224 for each @xmath353 . if @xmath19 is a frass structure , then tdc is further equivalent to the strong amalgamation property ( sap ) for the age of @xmath19 : for any @xmath1225 living on @xmath1226 respectively and embeddings @xmath1227 and @xmath1228 , there is @xmath1229 and embeddings @xmath1230 and @xmath1231 with @xmath1232 and @xmath1233 .
[ thm : afp ] let @xmath10 be a language and @xmath110 be a countably infinite @xmath10-structure .
the following are equivalent : 1 .
the logic action of @xmath122 on @xmath1234 ( @xmath93 the scott sentence of @xmath19 ) admits an invariant probability measure . 2 .
@xmath19 has tdc .
we will in fact need the following construction from ackerman - freer - patel s proof of [ thm : afp ] . starting with a countable @xmath10-structure @xmath19 with tdc
, they consider the morleyization @xmath1235 of the scott sentence @xmath93 of @xmath19 , where @xmath1236 with each @xmath1237 quantifier - free , as described following [ lm : einftys - smooth - gdelta ] .
they then produce ( see ( * ? ? ?
* section 3.4 ) ) a borel @xmath1097-structure @xmath1238 with universe @xmath89 such that for each @xmath353 and @xmath1239 , the corresponding subformula @xmath1240 in @xmath1241 is witnessed either by some @xmath1242 in the tuple @xmath1142 , or by all @xmath1242 in some nonempty open interval . clearly
then the restriction of @xmath1243 to any countable dense set of reals still satisfies @xmath1241 , hence ( its @xmath10-reduct ) is isomorphic to @xmath19 .
this shows : [ thm : afp - dense ] let @xmath10 be a language and @xmath19 be a countable @xmath10-structure with tdc .
then there is a borel @xmath10-structure @xmath1243 with universe @xmath89 such that for any countable dense set @xmath971 , @xmath1244 .
( marks ) if @xmath1 is smooth , then clearly it is @xmath19-structurable .
so we may assume @xmath1245 .
let @xmath1246 for @xmath1247 denote the basic clopen sets in @xmath228 .
note that the set @xmath1248_e
\cap n_s| = 1)\}\end{aligned}\ ] ] of points whose class contains an isolated point is borel , and @xmath1249 is smooth ( with a selector given by @xmath1250the unique element of @xmath4_e \cap n_s$ ] for the least @xmath1251 such that @xmath1252_e \cap n_s| = 1 $ ] ) , hence @xmath19-structurable . for @xmath1253 , the closure @xmath1254_e}$ ]
has no isolated points , hence is homeomorphic to @xmath228 . for each such @xmath164 , define @xmath1255 inductively for @xmath1256 by @xmath1257_e \cap n_{f_x(t ) } \subseteq n_s$ but $ [ x]_e \cap n_{f_x(t ) } \not\subseteq n_{s\^{\;\;}0 } , n_{s\^{\;\;}1}$}\end{aligned}\ ] ] ( for @xmath1258 ) , so that @xmath1259_e}$ ] , @xmath1260 is a homeomorphism , such that @xmath1261 .
it is easy to see that @xmath1262 is borel .
now let the structure @xmath1243 on @xmath89 be given by [ thm : afp - dense ] .
let @xmath1263 be a countable set so that there is a continuous bijection @xmath1264 .
let @xmath1265_e\ , ( f_x^{-1}(x ' ) \in z)\}.\end{aligned}\ ] ] then @xmath1266 is smooth ( with selector @xmath1267_e$ ] such that @xmath1268 with @xmath351 minimal ) , hence @xmath19-structurable .
finally , @xmath1269 is @xmath19-structurable : for each @xmath1270 , we have that @xmath1271_e ) \subseteq 2^\#n \setminus z$ ] is dense , so @xmath1272 gives a bijection between @xmath4_e$ ] and a dense subset of @xmath89 , along which we may pull back @xmath1243 to get a structure on @xmath4_e$ ] isomorphic to @xmath19 .
recently marks @xcite found , using baire category methods , another proof of [ thm : afp ] and of a generalization of [ thm : afp - dense ] , where @xmath89 can be replaced with any infinite polish space .
this allows a simplification in the above proof of [ thm : einftya - univ ] ( by omitting most of the second paragraph ) .
[ thm : einftya - univ ] has the following immediate corollary .
[ thm:817 ] the following frass structures can structure every aperiodic countable borel equivalence relation : @xmath102 , the random graph , the random @xmath103-free graph ( where @xmath103 is the complete graph on @xmath104 vertices ) , the random poset , and the rational urysohn space
. the concept of amenability of a structure in the next result can be either the one in @xcite ) or the one in @xcite .
this result was first proved by the authors by a different method but it can also be seen as a corollary of [ thm : einftya - univ ] .
[ thm : amenable - ntdc ] let @xmath19 be a countably infinite amenable structure
. then @xmath19 fails tdc .
since @xmath19 is amenable , every @xmath19-structurable equivalence relation is amenable ( see @xcite or @xcite ) , thus @xmath19 does not structure every aperiodic countable borel equivalence relation , and so @xmath19 fails tdc by [ thm : einftya - univ ] .
we do not know of a counterexample to the converse of [ thm : einftya - univ ] , i.e. , of a single structure @xmath19 without tdc such that every aperiodic @xmath69 is @xmath19-structurable .
such structures do exist which structure every compressible @xmath1 , as the following simple example shows : [ thm : qlex - univ ] for any countable linear order @xmath1273 , every compressible @xmath1274 is structurable via linear orders isomorphic to @xmath1275 with the lexicographical order . in particular , @xmath1276 structures every compressible equivalence relation .
by [ thm : einftya - univ ] , @xmath1 is structurable via linear orders isomorphic to @xmath214 .
take the lexicographical order on @xmath1277 .
concerning classes of structures ( or theories ) which can structure every ( compressible ) equivalence relation , we can provide the following examples : every @xmath300 is structurable via connected bipartite graphs .
the finite part of @xmath1 can be treed , so assume @xmath1 is aperiodic .
then we may partition @xmath1278 where @xmath1279 are complete sections ( this is standard ; see e.g. , @xcite ) .
let @xmath1280 and @xmath1281 be borel functions such that @xmath1282 and @xmath1283 for all @xmath1284 and @xmath1285 .
then the graph @xmath1286 which connects each @xmath1287 to @xmath1288 and @xmath1289 to @xmath1290 ( and with no other edges ) works . for every @xmath1291 ,
every compressible @xmath1274 is structurable via connected graphs in which all cycles have lengths divisible by @xmath664 .
let @xmath224 be a borel linear order on @xmath2 , and let @xmath1286 be any borel graphing , e.g. , @xmath1292 .
let @xmath1293 let @xmath957 be the equivalence relation on @xmath946 generated by @xmath1 and @xmath1294 for @xmath334 , @xmath1295 , and @xmath1296 , and let @xmath1297 be the graph generated by , for each @xmath1298 with @xmath1299 , @xmath1300 ( and no other edges ) .
that is , every edge in @xmath240 has been replaced by a @xmath664-length path with the same endpoints .
it is clear that @xmath1048 graphs @xmath957 and every cycle in @xmath1048 has length divisible by @xmath664 .
now since @xmath1 is compressible , and the inclusion @xmath1301 is a complete section embedding @xmath1302 , we have @xmath958 , thus @xmath1 is structurable via a graph isomorphic to @xmath1048 .
thus in contrast to the fact that not every ( compressible ) countable borel equivalence relation is treeable , we have every compressible equivalence relation has graphings of arbitrarily large girth .
we do not know if this holds for arbitrary countable borel equivalence relations .
let us say that an elementary class @xmath66 is scott axiomatizable if it is axiomatizable by a scott sentence @xmath93 of some structure @xmath19 , or equivalently by some sentence @xmath34 which is countably categorical ( i.e. , it has exactly one countable model up to isomorphism ) .
several elementary classes we have considered are naturally scott axiomatizable : e.g. , aperiodic , aperiodic smooth , aperiodic hyperfinite ( by @xmath1303 ) , free actions of a group @xmath21 ( by @xmath288 ) , and compressible ( by the sentence in the language @xmath1304 asserting that @xmath23 is the graph of an injective function with infinite and coinfinite image and with no fixed points ) .
it is an open problem to characterize the elementary classes which are scott axiomatizable .
in fact , we do not even know if every elementary class of aperiodic equivalence relations is scott axiomatizable . here
we describe a general construction which can be used to show that certain compressible elementary classes are scott axiomatizable .
let @xmath1305 be theories .
let @xmath1306 be a sentence in the language @xmath1307 asserting 1 .
@xmath1308 are equivalence relations such that the quotient maps @xmath1309 and @xmath1310 ( where @xmath2 is the universe ) exhibit a bijection between @xmath2 and @xmath1311 ; and 2 .
the @xmath10-reduct ( respectively @xmath1312-reduct ) is an @xmath1313-invariant ( resp .
, @xmath1314-invariant ) structure which induces a model of @xmath34 ( resp . , @xmath816 ) on the quotient @xmath1315 ( resp . , @xmath1316 ) .
thus , a @xmath1317-structure @xmath19 on a set @xmath2 is essentially the same thing as a @xmath34-structure @xmath312 on a set @xmath118 and a @xmath816-structure @xmath1318 on a set @xmath311 , together with a bijection @xmath1319 .
the following are clear : @xmath1320 ( equivalently , @xmath1321 ) .
it is not true in general that @xmath1322 .
for example , if @xmath1323 and @xmath1324 axiomatize free actions of countable groups @xmath1325 , then it is easy to see that @xmath1326 axiomatizes free actions of @xmath1327 ; taking @xmath1328 , we have that @xmath1329 is the universal orbit equivalence of a free action of @xmath1330 , which does not reduce to a product of two treeables ( such as @xmath1331 ) by @xcite ) .
if @xmath108 are countably categorical , then so is @xmath1306 .
now consider the case where @xmath816 in the language @xmath1332 asserts that the @xmath98 are disjoint singleton subsets which enumerate the universe .
then clearly @xmath816 axiomatizes the aperiodic smooth countable borel equivalence relations , i.e. , @xmath1333 , whence @xmath1334 . for this choice of @xmath816 , @xmath1335 .
let @xmath1329 live on @xmath2 and let @xmath1336 . then from the definition of @xmath1306 , we have that ( the reduct to the language of @xmath34 of ) @xmath1337 ( where @xmath98 is from the language of @xmath816 as above ) .
let @xmath1338 .
then it is easy to see that @xmath1339 , where @xmath1340 for the unique @xmath353 such that @xmath1341
. since @xmath816 is clearly countably categorical , this yields if an elementary class @xmath451 is scott axiomatizable , then so is @xmath1342 .
in particular , if an elementary class @xmath39 is scott axiomatizable and closed under @xmath730 , then @xmath1343 ( i.e. , the compressible elements of @xmath39 ) is scott axiomatizable . if @xmath1344 where @xmath34 is countably categorical , then @xmath1345 ( using [ thm : einfty - e - in ] ) , whence @xmath1346 . for the second statement , if @xmath448 where @xmath1 is universally structurable , then @xmath1347 ( [ thm : compress - univstr ] ) .
[ thm : compress - hyperfinite - treeable - scott ] the following elementary classes are scott axiomatizable : compressible hyperfinite , compressible treeable . for the compressible treeables , use that @xmath1348 ( i.e. , the @xmath220-universal orbit equivalence of a free action of @xmath218 ) is @xmath186-universal treeable @xcite ; it follows that @xmath1349 is closed under @xmath730 , and also that @xmath1350 is the class of compressible treeables .
however , we do not know if the elementary class of aperiodic treeable equivalence relations is scott axiomatizable .
at the end of [ sec : tensor ] we asked : is @xmath564 universally structurable ( or equivalently , isomorphic to @xmath456 ) for every aperiodic @xmath1 ?
the following question ( [ rmk : aperiodic - univstr - compress ] ) concerns the structure of universally structurable @xmath188-classes : is @xmath743 for every aperiodic universally structurable @xmath1 ?
equivalently , is the compressible element of every universally structurable @xmath188-class the @xmath220-least of the aperiodic elements ? by [ thm : pbr ] , we know that there are many incomparable elementary reducibility classes , or equivalently , many @xmath86-incomparable universally structurable @xmath1 .
however , these were produced using the results in @xcite , which use rigidity theory for measure - preserving group actions .
one hope for the theory of structurability is the possibility of producing @xmath86-incomparable equivalence relations using other methods , e.g. , using model theory .
show that there are @xmath86-incomparable @xmath1351 without using ergodic theory .
we turn now to the order - theoretic structure of the lattice @xmath82 ( equivalently , the poset of elementary classes ) and the lattice @xmath88 ( equivalently , the poset of elementary reducibility classes ) .
the following questions , posed in [ sec : lattice ] ( near end ) , are natural from an abstract order - theoretic perspective , though perhaps not so approachable : [ prb : lattice - complete ] is either @xmath82 or @xmath88 a complete lattice ?
if so , is it completely distributive ? is either @xmath82 or @xmath88 a zero - dimensional @xmath83-complete lattice , in that it embeds into @xmath913 for some set @xmath2 ?
we noted above ( [ thm : lattice-0d ] ) that the recent work of marks @xcite gives some examples of @xmath83-prime filters on @xmath88 , and also ( [ thm : indepjoin - treeable ] ) that these filters can not separate elements of @xmath88 below the universal treeable equivalence relation @xmath1352 . regarding [ prb : lattice - complete ] , a natural attempt at a negative answer would be to show that some `` sufficiently complicated '' collection of universally structurable equivalence relations does not have a join .
for example , one could try to find the join of a strictly increasing @xmath83-sequence .
[ prb : omega1-seq ] is there an `` explicit '' strictly increasing @xmath83-sequence in @xmath82 ?
similarly for @xmath1353 .
note that by [ thm : pbr ] , such a sequence does exist , abstractly ; the problem is thus to find a sequence which is in some sense `` definable '' , preferably corresponding to some `` natural '' hierarchy of countable structures .
for example , a long - standing open problem ( implicit in e.g. , ( * ? ? ?
* section 2.4 ) ) asks whether the sequence of elementary classes @xmath1354 , where @xmath1355 stabilizes ( or indeed is constant ) ; a negative answer would constitute a positive solution to [ prb : omega1-seq ] .
one possible approach to defining an @xmath83-sequence would be by iterating a `` jump '' operation , @xmath1356 , that sends any non - universal @xmath1357 to some non - universal @xmath1358 such that @xmath1359 .
is there an `` explicit '' jump operation on the non - universal elements of @xmath1353 ? on the other hand , this would not be possible if there were a greatest non - universal element : is there a greatest element among the non - universal elements of @xmath911 , or of @xmath1353 ?
if so , do the non - universal equivalence relations form an elementary class , i.e. , are they downward - closed under @xmath187 ? the general model - theoretic question concerning structurability is which properties of a theory @xmath109 ( or a borel class of structures @xmath0 ) yield properties of the corresponding elementary class @xmath40 ( or @xmath1360 ) . fits into this mold , by characterizing the @xmath34 which yield smoothness .
one could seek similar results for other properties of countable borel equivalence relations .
find a model - theoretic characterization of the @xmath34 such that @xmath40 consists of only hyperfinite equivalence relations , i.e. , such that @xmath1361 , for any sentence @xmath1362 axiomatizing hyperfiniteness .
less ambitiously , one might look for `` natural '' examples of such @xmath34 , for specific classes of structures .
for example , for the borel class of locally finite graphs , we have : * if @xmath1 is structurable via locally finite trees with one end , then @xmath1 is hyperfinite @xcite . * if @xmath1 is structurable via locally finite graphs with two ends , then @xmath1 is hyperfinite @xcite .
if @xmath1 is structurable via locally finite graphs with at least 3 but finitely many ends , then @xmath1 is smooth ; this follows from @xcite , or simply by observing that in any such graph , a finite nonempty subset may be defined as the set of all vertices around which the removal of a ball of minimal radius leaves @xmath1363 infinite connected components .
find `` natural '' examples of @xmath34 such that @xmath40 consists of only frchet - amenable equivalence relations ( see @xcite ) .
for example , every @xmath1 structurable via countable scattered linear orders is frchet - amenable @xcite ( recall that a countable linear order is scattered if it does not embed the rationals ) ; note however that the scattered linear orders do not form a borel class of structures .
find `` natural '' examples of @xmath34 such that @xmath40 consists of only compressible equivalence relations .
for example , by @xcite , the class of @xmath1 structurable via locally finite graphs whose space of ends is not perfect but has cardinality at least 3 is exactly @xmath738 .
there is also the converse problem of determining for which @xmath34 is every equivalence relation of a certain form ( e.g. , compressible ) @xmath34-structurable .
fits into this mold , by giving a sufficient condition for a single structure to structure every aperiodic equivalence relation .
is there a structure @xmath19 without tdc which structures every aperiodic countable borel equivalence relation ?
that is , does the converse of [ thm : einftya - univ ] hold ?
in particular , does @xmath1276 structure every aperiodic equivalence relation ?
we noted above that it structures every compressible equivalence relation , thus the analogous question for the compressibles has a negative answer . find a model - theoretic characterization of the structures @xmath19 such that every compressible equivalence relation is @xmath19-structurable , i.e. , @xmath1364 .
we also have the corresponding questions for theories ( or classes of structures ) : find a model - theoretic characterization of the @xmath34 such that @xmath1365 , or more generally , @xmath1366 .
we gave several examples in [ sec : einftys - univ ] .
another example is the following @xcite : every @xmath69 is structurable via locally finite graphs with at most one end . for a different sort of property that @xmath40 may or may not have , recall ( [ sec : elemred ] ) that @xmath40 is an elementary reducibility class , i.e. , closed under @xmath86 , when @xmath34 axiomatizes linear orders embeddable in @xmath20 , or when @xmath34 axiomatizes trees .
find a model - theoretic characterization of the @xmath34 such that @xmath40 is closed under @xmath86 .
we considered in [ sec : scott ] the question of which elementary classes are scott axiomatizable , i.e. , axiomatizable by a scott sentence .
find other `` natural '' examples of scott axiomatizable elementary classes .
we showed above ( [ thm : compress - hyperfinite - treeable - scott ] ) that the class of compressible treeable equivalence relations is scott axiomatizable .
is the class of aperiodic treeable ( countable borel ) equivalence relations scott axiomatizable
? the class of aperiodic treeables can not be axiomatized by the scott sentence of a single countable tree @xmath329 . indeed , since @xmath250 would have to be treeable by @xmath329 , by a result of adams ( see @xcite ) , @xmath329 can have at most @xmath219 ends ; but then by @xcite and @xcite , every @xmath1 treeable by @xmath329 is hyperfinite .
find a model - theoretic characterization of the @xmath34 such that @xmath40 is axiomatizable by a scott sentence ( possibly in some other language ) .
in particular , is every elementary class of aperiodic , or compressible , equivalence relations axiomatizable by a scott sentence ?
we conclude by stating two very general ( and ambitious ) questions concerning the relationship between structurability and model theory . for the first , note that by ( i)@xmath1367(ii ) in [ thm : einftys - smooth ] ,
the condition @xmath1060 is equivalent to the existence of a formula(s ) in the language of @xmath34 with some definable properties which are logically implied by @xmath34 .
our question is whether a similar equivalence continues to hold when @xmath1059 is replaced by an arbitrary sentence @xmath816 .
[ prb : elem - interp ] is there , for any @xmath816 , a sentence @xmath1368 in a language consisting of relation symbols @xmath1369 ( thought of as `` predicate variables '' ) , such that for any @xmath34 , we have @xmath815 iff there are formulas @xmath1370 in the language of @xmath34 such that @xmath34 logically implies @xmath1371 ( the result of `` substituting '' @xmath1372 for @xmath8 in @xmath1373 ) ? finally , there is the question of completely characterizing containment between elementary classes : find a model - theoretic characterization of the pairs @xmath1374 such that @xmath815 .
in this appendix , we discuss fiber spaces on countable borel equivalence relations , which provide a more general context for structurability and related notions .
the application of fiber spaces to structurability was previously considered in @xcite and ( * ? ? ?
* appendix d ) . in both this appendix and
the next , we will use categorical terminology somewhat more liberally than in the body of this paper .
let @xmath300 be a countable borel equivalence relation .
a fiber space over @xmath1 consists of a countable borel equivalence relation @xmath1375 , together with a surjective countable - to-1 class - bijective homomorphism @xmath1376 .
we refer to the fiber space by @xmath1377 , by @xmath1378 , by @xmath1379 , or ( ambiguously ) by @xmath776 .
we call @xmath1375 the total space , @xmath133 the base space , and @xmath1023 the projection . for @xmath163 ,
the fiber over @xmath164 is the set @xmath1380 .
for @xmath1381 such that @xmath1382 , we let @xmath1383 denote the fiber transport map , where for @xmath1384 , @xmath1385 is the unique @xmath1386 such that @xmath1387 . for two fiber spaces
@xmath1388 over @xmath133 , a fiberwise map between them over @xmath1 , denoted @xmath1389 ( we use letters like @xmath1390 for maps between total spaces ) , is a homomorphism @xmath1391 such that @xmath1392 ( note that this implies that @xmath321 is class - bijective ) : @xmath1393 ( u , p ) \drar[swap]{p } \ar{rr}{\~f } & & ( v , q ) \dlar{q } \\ & ( x , e ) \end{tikzcd}\ ] ] for a fiber space @xmath1377 over @xmath133 and a fiber space @xmath1394 over @xmath169 , a fiber space homomorphism from @xmath1378 to @xmath1395 , denoted @xmath1396 , consists of two homomorphisms @xmath655 and @xmath1391 such that @xmath1397 : @xmath1398{p } \rar{\~f } & ( v , q ) \dar{q } \\ ( x , e ) \rar[swap]{f } & ( y , f ) \end{tikzcd}\ ] ] we sometimes refer to @xmath321 as the fiber space homomorphism ; note that @xmath47 is determined by @xmath321 ( since @xmath1023 is surjective ) .
we say that @xmath321 is a fiber space homomorphism over @xmath47 .
note that a fiberwise map over @xmath1 is the same thing as a fiber space homomorphism over the identity function on @xmath1 .
a fiber space homomorphism @xmath1396 is fiber - bijective if @xmath1399 is a bijection for each @xmath163 ( where @xmath1 lives on @xmath2 ) ; fiber - injective , fiber - surjective are defined similarly .
let @xmath1377 be a fiber space over @xmath133 , and let @xmath394 be a countable borel equivalence relation with a homomorphism @xmath1400 .
recall ( [ sec : pullback ] ) that we have the fiber product equivalence relation @xmath1401 with respect to @xmath47 and @xmath1023 , which comes equipped with the canonical projections @xmath1402 and @xmath1403 obeying @xmath1404 .
it is easy to check that @xmath247 is class - bijective , surjective , and countable - to-1 ( because @xmath1023 is ) . in this situation
, we also use the notation @xmath1405 note that @xmath1406 is then a fiber space homomorphism over @xmath47 .
we refer to @xmath1407 as the pullback of @xmath1377 along @xmath47 .
here is a diagram : @xmath1408{f^{-1}(p ) } \rar{\~f } & ( u , p ) \dar{p } \\
( y , f ) \rar[swap]{f } & ( x , e ) \end{tikzcd}\ ] ] let @xmath1409 denote the category of fiber spaces and fiberwise maps over @xmath1 , and let @xmath1410 denote the category of fiber spaces and fiber space homomorphisms . for a homomorphism @xmath655
, pullback along @xmath47 gives a functor @xmath1411 ( with the obvious action on fiberwise maps ) .
the assignment @xmath1412 is itself functorial , and turns @xmath1413 into a contravariant functor from the category @xmath249 to the category of ( essentially small ) categories .
( technically @xmath1412 is only _ _
pseudo__functorial , i.e. , @xmath1414 is naturally isomorphic , not equal , to @xmath1415 ; we will not bother to make this distinction . )
let @xmath1377 be a fiber space over @xmath300 .
by lusin - novikov uniformization , we may borel partition @xmath2 according to the cardinalities of the fibers .
suppose for simplicity that each fiber is countably infinite .
again by lusin - novikov uniformization , there is a borel map @xmath1416 such that each @xmath327 is a bijection @xmath1417 .
let @xmath340 be the cocycle given by @xmath1418 ( where @xmath1419 is the fiber transport map ; compare [ rmk : skew - product ] ) .
we then have a ( fiberwise ) isomorphism of fiber spaces over @xmath1 , between @xmath1377 and the skew product @xmath1420 ( with its canonical projection @xmath1421 ) : @xmath1422 recall that two cocycles @xmath1423 are cohomologous if there is a borel map @xmath1424 such that @xmath1425 , for all @xmath1426 . it is easy to see that in the above , changing the map @xmath1416 results in a cohomologous cocycle @xmath340 ; so we get a well - defined map from ( isomorphism classes of ) fiber spaces over @xmath1 with countably infinite fibers to @xmath343-valued cohomology classes on @xmath1 .
conversely , given any cocycle @xmath1427 , the skew product @xmath1428 yields a fiber space over @xmath1 with countably infinite fibers .
these two operations are inverse to each other , so we have a bijection @xmath1429 in fact , we have the following more refined correspondence , which also smoothly handles the case with finite fibers .
let @xmath1430 denote the category whose objects are @xmath1431 and morphisms are maps between them ( where as usual , @xmath1432 for @xmath253 ) . then @xmath1430 is a `` standard borel category '' . regarding @xmath1 as the groupoid on @xmath2 with a single morphism between any two related points , we have a borel functor category @xmath1433 , whose objects are borel functors @xmath1434 and morphisms are borel natural transformations .
we then have a functor @xmath1435 which takes a borel functor @xmath1436 to the obvious generalization of the skew product of @xmath1 with respect to @xmath275 ( but where the fibers are no longer uniformly @xmath325 , but vary from point to point according to @xmath275 ) ; and this functor is an equivalence of categories . we leave the details to the reader . using this correspondence between fiber spaces and cocycles ,
we obtain [ thm : pinfty ] there is a fiber space @xmath1437 over @xmath221 , which is universal with respect to fiber - bijective invariant embeddings : for any other fiber space @xmath1377 over @xmath1 , there is a fiber - bijective homomorphism @xmath1438 over an invariant embedding @xmath1439 . for simplicity , we restrict again to the case where @xmath776 has countably infinite fibers .
let @xmath34 be a sentence over the language @xmath1440 , where each @xmath1441 is binary , asserting that @xmath1442 defines a cocycle @xmath1443 , where @xmath2 is the universe of the structure ( and @xmath1444 is the indiscrete equivalence relation on @xmath2 ) .
then the canonical @xmath34-structure on @xmath62 corresponds to a cocycle @xmath1445 .
we will in fact define the universal fiber space @xmath1446 over @xmath62 , since clearly @xmath1447 ( by giving @xmath221 the trivial cocycle ) .
let @xmath1448 , with @xmath1449 the canonical projection .
for another fiber space @xmath1377 over @xmath1 with countably infinite fibers , by the above remarks , @xmath776 is isomorphic ( over @xmath1 ) to a skew product @xmath1428 , for some cocycle @xmath1427 .
this @xmath275 corresponds to a @xmath34-structure on @xmath1 , which yields an invariant embedding @xmath1450 such that @xmath275 is the restriction of @xmath1451 along @xmath47 , giving the desired fiber - bijective homomorphism @xmath1452 over @xmath47 .
there is a different kind of universality one could ask for , which we do not know how to obtain .
namely , for each @xmath69 , is there a fiber space @xmath1437 over @xmath1 which is universal with respect to fiberwise injective maps ?
let @xmath300 be a countable borel equivalence relation .
the tautological fiber space over @xmath1 is @xmath1453 , where @xmath1454 is the equivalence relation on the set @xmath1455 given by @xmath1456 and @xmath1457 is the first coordinate projection ( i.e. , @xmath1458 ) . in other words ,
the @xmath1454-fiber over each @xmath1-class @xmath235 consists of the elements of @xmath16 .
note that @xmath1454 is the kernel of the second coordinate projection @xmath1459 ; thus @xmath1454 is smooth , and in fact @xmath1460 is isomorphic to @xmath2 ( via @xmath989 ) .
now let @xmath1377 be any smooth fiber space over @xmath133 , and let @xmath43 be the ( countable borel ) equivalence relation on @xmath1461 given by @xmath1462_p \mathrel{f } [ u']_p \iff p(u ) \mathrel{e } p(u').\end{aligned}\ ] ] by lusin - novikov uniformization , there is a borel map @xmath1463 which is a section of @xmath1023 , which when composed with the projection @xmath1464 gives a reduction @xmath158 whose image is a complete section .
let us say that a presentation of the quotient space @xmath1465 consists of a countable borel equivalence relation @xmath394 together with a bijection @xmath1466 which admits a borel lifting @xmath1467 ( which is then a reduction @xmath171 with image a complete section ) .
by the above , every smooth fiber space over @xmath1 gives rise to a presentation of @xmath1465 .
conversely , given any presentation @xmath169 of @xmath1465 , letting @xmath701 with image a complete section , the pullback @xmath1468 is a fiber space over @xmath1 , which is smooth ( because @xmath1468 reduces to @xmath1469 , via the map @xmath321 coming from the pullback ) .
it is easily seen that the two operations we have just described are mutually inverse up to isomorphism , yielding a bijection @xmath1470 this correspondence between smooth fiber spaces and presentations of the same quotient space is essentially the proof of ( * ? ? ?
* d.1 ) .
we now describe the correspondence between homomorphisms of equivalence relations and fiber space homomorphisms .
let @xmath153 .
a homomorphism @xmath655 induces a fiber space homomorphism @xmath1471 over @xmath47 , given by @xmath1472 conversely , let @xmath1473 be any fiber space homomorphism over @xmath1474 .
then @xmath595 must be given by @xmath1475 for some @xmath116 such that @xmath1476 for each @xmath163 ; in particular , @xmath47 is a homomorphism @xmath1477 .
let us say that two homomorphisms @xmath1478 are equivalent , denoted @xmath1479 , if @xmath1480 for each @xmath163 ; equivalently , they induce the same map on the quotient spaces @xmath157 .
the above yield mutually inverse bijections @xmath1481 class - injectivity on the left translates to fiber - injectivity on the right , etc .
we discuss here an alternative point of view on fiber spaces and equivalence relations .
the idea is that the tautological fiber space @xmath1454 over an equivalence relation @xmath133 allows a clean distinction to be made between the quotient space @xmath1465 and the presentation @xmath133 .
a countable borel quotient space is , formally , the same thing as a countable borel equivalence relation @xmath133 , except that we denote it by @xmath1465 .
a borel map between countable borel quotient spaces @xmath1465 and @xmath1482 , denoted @xmath1483 , is a map which admits a borel lifting @xmath1484 , or equivalently an @xmath1485-class of borel homomorphisms @xmath1477 .
let @xmath1486 denote the category of countable borel quotient spaces and borel maps .
( note that @xmath1487 are isomorphic in @xmath1486 iff they are _ bireducible _ as countable borel equivalence relations . )
let @xmath1488 denote the class of standard borel spaces . by identifying @xmath1489 with @xmath1490
, we regard @xmath1491 as a full subcategory of @xmath1486 . by regarding borel maps in @xmath1492 as @xmath1485-classes of homomorphisms
, we have that @xmath1486 is the quotient category of @xmath249 ( with the same objects ) by the congruence @xmath1485 . a ( quotient ) fiber space over a quotient space @xmath1493 is a quotient space @xmath1494 together with a countable - to-1 surjection @xmath1495 .
this definition agrees with the previous notion of fiber space over @xmath133 , in that fiber spaces over @xmath1465 are in natural bijection with fiber spaces over @xmath133 , up to isomorphism .
indeed , by [ thm : smh - factor ] , we may factor any lifting @xmath1496 of @xmath1023 into a reduction with image a complete section , followed by a class - bijective homomorphism ; the former map becomes an isomorphism when we pass to the quotient , so @xmath1497 is isomorphic to a fiber space with class - bijective projection .
we have obvious versions of the notions of fiberwise map over @xmath1465 , fiber space homomorphism , and fiber - bijective homomorphism for quotient fiber spaces .
let @xmath1498 denote the category of fiber spaces over @xmath1465 ; in light of the above remarks , @xmath1498 is equivalent to @xmath1409 .
let @xmath1499 denote the category of quotient fiber spaces and homomorphisms ( @xmath1499 is then the quotient of @xmath1410 by @xmath1485 ) .
we now have a full embedding @xmath1500 that sends an equivalence relation @xmath133 to its tautological fiber space @xmath1501 but regarded as the quotient fiber space @xmath1502 over @xmath1465 , and sends a homomorphism @xmath47 to the corresponding fiber space homomorphism @xmath1503 given above .
thus , we may regard equivalence relations as special cases of fiber spaces over quotient spaces .
to summarize , here is a ( non - commuting ) diagram of several relevant categories and functors : @xmath1504 \\
( \@e , ->_b ) \rar[hook ] \drar[- > > ] & \int_\@q \*{fib } \dar \\
( \@b , ->_b ) \rar[hook ] & ( \@q , ->_b ) \end{tikzcd}\ ] ] the horizontal arrows are full embeddings , the diagonal arrows are quotients by @xmath1485 , and the vertical arrows are forgetful functors that send a fiber space to its base space .
let @xmath1388 be fiber spaces over @xmath190 respectively , and @xmath655 .
a fiber space homomorphism @xmath1505 over @xmath47 corresponds , via the universal property of the pullback @xmath1506 , to a fiberwise map @xmath1507 over @xmath1 : @xmath1508{ddr}[swap]{p } \drar[swap , pos=.8]{\~f ' } \ar{drr}{\~f } \\[-1em ] ~ & f^{-1}(q ) \dar{f^{-1}(q ) } \rar & q \dar{q } \\ ~ & e \rar[swap]{f } & f \end{tikzcd}\ ] ] note that @xmath321 is fiber - bijective iff @xmath1509 is an isomorphism . in general , since @xmath1509 is countable - to-1 , we may further factor it into the surjection onto its image @xmath1510 followed by an inclusion : @xmath1511{ddrr}[swap]{p } \drar[->>][swap , pos=.8]{\~f ' } \ar[bend left=5]{drrr}{\~f } \\[-1em ] ~ & \~f'(p ) \drar \rar[hook ] & [ -1em ] f^{-1}(q ) \dar{f^{-1}(q ) } \rar & q \dar{q } \\ ~ & ~ & e \rar[swap]{f } & f \end{tikzcd}\ ] ] so we have a canonical factorization of any fiber space homomorphism @xmath321 into a fiberwise surjection over @xmath1 , followed by a fiberwise injection over @xmath1 , followed by a fiber - bijective homomorphism . in the case where @xmath1512 , @xmath1513 , and @xmath1514 is the fiber space homomorphism induced by @xmath47
, the fiber space @xmath1515 is given by @xmath1516 while the map @xmath1517 is given by @xmath1518 comparing with the proofs of [ thm : classinj - factor , thm : smh - factor ] reveals that when @xmath655 is smooth , the above factorization of @xmath1503 corresponds ( via the correspondence between smooth fiber spaces and presentations from [ sec : fiber - equiv ] ) to the factorization of @xmath47 produced by [ thm : smh - factor ] .
in particular , we obtain a characterization of smooth homomorphisms in terms of fiber spaces : @xmath655 is smooth iff the fiber space @xmath1468 over @xmath1 is smooth .
in fact , the proof of [ thm : classinj - factor ] is essentially just the above correspondence , plus the observation that @xmath1519 and smoothness of @xmath1469 imply that @xmath1520 is smooth ( compare also ( * ? ? ? * d.2 ) ) .
let @xmath10 be a language and @xmath1377 be a fiber space over @xmath300 .
a borel @xmath10-structure on @xmath1377 is a borel @xmath10-structure @xmath1521 with universe @xmath1522 which only relates elements within the same fiber , i.e. , @xmath1523 such that structures on fibers over the same @xmath1-class are related via fiber transport , i.e. , @xmath1524 for an @xmath33-sentence @xmath34 , we say that @xmath19 is a borel @xmath34-structure on @xmath1377 , denoted @xmath1525 if @xmath1526 satisfies @xmath34 for each @xmath163 . for @xmath300 ,
@xmath34-structures on @xmath1 are in bijection with @xmath34-structures on the tautological fiber space @xmath1527 over @xmath1 , where @xmath1073 corresponds to @xmath1528 given by @xmath1529 in other words , for each @xmath163 , @xmath1530_e$ ] and @xmath1531 are isomorphic via the canonical bijection @xmath1532 between @xmath4_e$ ] and @xmath1533 . for a fiber space homomorphism @xmath1396 and a @xmath34-structure @xmath1534 , the fiberwise pullback structure @xmath1535
is defined in the obvious way , i.e. , @xmath1536 we have the following generalization of [ thm : esigma ] : let @xmath1377 be a fiber space over @xmath300 and @xmath109 be a theory .
there is a fiber space @xmath1537 over an equivalence relation @xmath1538 , together with a fiber - bijective homomorphism @xmath1539 and a @xmath34-structure @xmath1540 , such that the triple @xmath1541 is universal : for any other fiber space @xmath1394 over @xmath394 with a fiber - bijective homomorphism @xmath1542 and a structure @xmath1534 , there is a unique fiber - bijective @xmath1543 such that @xmath1544 and @xmath1545 .
this is a straightforward generalization of [ thm : esigma ] ( despite the excessive notation ) .
the equivalence relation @xmath1546 lives on @xmath1547 and is given by @xmath1548 as usual , the borel structure on @xmath1546 is given by uniformly enumerating each @xmath964 .
the base space part of @xmath314 is given by @xmath1549 , the fiber space @xmath1550 is given by the pullback @xmath1551 , and the structure @xmath316 is given by @xmath1552 for @xmath163 , @xmath1553 , and @xmath1554 .
the universal property is straightforward .
however , our other basic universal construction for structuring equivalence relations , the `` scott sentence '' ( [ thm : sigmae ] ) , fails to generalize in a straightforward fashion to fiber spaces ; this is essentially because we require languages to be countable , whereas the invariant borel @xmath34-algebra of a nonsmooth fiber space is not countably generated .
nonetheless , we may define the fiber - bijective product @xmath1555 of two fiber spaces @xmath1388 over @xmath153 respectively , by generalizing [ thm : tensor - alt ] , yielding their categorical product in the category of fiber spaces and fiber - bijective homomorphisms ; we leave the details to the reader .
in particular , by taking @xmath1395 to be the universal fiber space @xmath1556 from [ thm : pinfty ] , we obtain for every fiber space @xmath1379 over @xmath69 , there is a fiber space @xmath1557 admitting a fiber - bijective homomorphism to @xmath1378 and which is universal among such fiber spaces with respect to fiber - bijective embeddings .
we conclude by noting that restricting attention to smooth fiber spaces and applying the correspondence with presentations gives a different perspective on some results from [ sec : reductions ] : * ( * ? ? ?
* d.1 ) if @xmath300 admits a smooth fiber space @xmath1379 , and @xmath1558 , then @xmath1379 corresponds to a presentation @xmath169 of @xmath1465 , and @xmath19 corresponds to a structure @xmath1559 , i.e. , a structure @xmath65 ; hence @xmath1 is bireducible with a @xmath34-structurable equivalence relation . * in particular , if @xmath599 and @xmath305 , then pulling back @xmath1560 along @xmath47 gives a smooth @xmath34-structured fiber space ( namely @xmath1468 ) over @xmath1 , whence @xmath1 is bireducible with a @xmath34-structurable equivalence relation .
so @xmath931 is closed under @xmath609 ( [ thm : red - elem ] ) . *
if @xmath614 , then the induced @xmath1471 is fiber - injective , which yields a fiberwise injection @xmath1561 over @xmath1 , whence @xmath1 embeds into the @xmath34-structurable presentation corresponding to @xmath1468 ; this similarly re - proves part of [ thm : emb - elem ] .
we discuss here a categorical structure on the class of all theories , which interacts well with several of the constructions we have considered .
let @xmath1562 denote the class of all theories .
let @xmath1563 be theories .
we denote logical equivalence of @xmath33-formulas by @xmath1564 , and logical equivalence modulo @xmath34 by @xmath1565 . by an interpretation of @xmath10 in @xmath1566 , written @xmath1567 we mean a function @xmath1568 mapping each @xmath104-ary relation @xmath111 to a @xmath816-equivalence class of @xmath1569-formulas with free variables from @xmath1570 .
we generally abuse notation by identifying equivalence classes of formulas with individual formulas .
thus , for @xmath111 , we denote by @xmath1571 any formula in the equivalence class @xmath1572 . given an interpretation @xmath1573 and an @xmath33-formula @xmath1574
, we define the @xmath1569-formula @xmath1575 ( modulo @xmath1576 ) by `` substituting '' @xmath1572 for @xmath23 in @xmath1064 , for each @xmath111 ; formally , this is defined by induction on @xmath1064 in the obvious manner .
( in general this will require renaming bound variables in @xmath1064 ; hence @xmath1577 is only well - defined modulo @xmath1576 . ) given an interpretation @xmath1573 and a model @xmath1578 of @xmath816 , the @xmath275-reduct of @xmath19 is the @xmath10-structure @xmath1579 given by @xmath1580 it follows by induction that for any @xmath33-formula @xmath1064 , we have @xmath1581 .
when @xmath275 is the inclusion of a sublanguage @xmath1582 , @xmath1583 is the @xmath10-reduct in the usual sense . now by an interpretation of @xmath109 in @xmath1566 , written @xmath1584 we mean an interpretation @xmath275 of @xmath10 in @xmath1566 such that the @xmath1569-sentence @xmath1585 is logically implied by @xmath816 ; equivalently , for any ( countable ) model @xmath1586 , we have @xmath1587 . this notion of `` interpretation '' is more restrictive than the usual notion considered in model theory ( see ( * ? ? ?
* section 5.3 ) ) .
given interpretations @xmath1588 and @xmath1589 , we may compose them in the obvious manner to get @xmath1590 , where @xmath1591 for @xmath111 .
we also have the identity interpretation @xmath1592 .
thus , we have a category @xmath1593 of theories and interpretations .
the category @xmath1593 has an initial object @xmath1594 where @xmath1595 is a tautology , as well as a terminal object @xmath1596 where @xmath771 is a contradictory sentence .
it also has countable coproducts and products , given by the operations @xmath822 and @xmath821 respectively ( _ not _ vice - versa ) from [ sec : lattice ] ; we sketch here the verification of the universal properties .
let @xmath825 be a countable family of theories .
we verify that their coproduct is @xmath1597 with the canonical injections @xmath1598 given by the inclusions @xmath1599 .
let @xmath1566 be another theory , and let @xmath1600 be interpretations for each @xmath353 ; we must find a unique interpretation @xmath1601 such that @xmath1602 for each @xmath353 .
this condition means precisely that @xmath1603 for each @xmath1604 , which determines @xmath275 uniquely as an interpretation @xmath1605 ; and since @xmath816 logically implies each @xmath1606 , @xmath816 also logically implies @xmath1607 , whence @xmath275 is an interpretation @xmath1608 , as desired .
now we verify that the product is @xmath1609 with the canonical projections @xmath1610 given by @xmath1611 a computation shows that @xmath1612 is logically equivalent to ( hence logically implied by ) @xmath834 , whence @xmath1613 is an interpretation @xmath1614 . by a straightforward induction , 1 . for each @xmath1615-formula @xmath1064 , we have @xmath1616 .
now let @xmath1566 be another theory , and let @xmath1617 for each @xmath353 ; we must find a unique @xmath1618 such that @xmath1619 for each @xmath353 .
this condition means that for each @xmath1620 and each @xmath353 , we have @xmath1621 .
so by ( @xmath398 ) , we must have @xmath1622 this determines @xmath275 uniquely . to check that @xmath851 logically implies @xmath1623 , use ( @xmath398 ) on @xmath1623 , together with @xmath1624 ( by definition of @xmath275 and @xmath1613 )
; we omit the details . in [ sec : interp - ooobool ] we will give a more abstract description of countable coproducts ( and other countable colimits ) in @xmath1593 .
we now relate interpretations to structurability . for an interpretation @xmath1588 and a countable borel equivalence relation @xmath300 with a @xmath816-structure @xmath1625 ,
the classwise @xmath275-reduct of @xmath19 is the @xmath34-structure @xmath1626 where @xmath1627_e}(\-x)\end{aligned}\ ] ] for @xmath104-ary @xmath111 and @xmath1132 with @xmath318 . in other words , for each @xmath1-class @xmath235 , we have @xmath1628 ( note that in general , @xmath1629 . ) clearly , @xmath1630 preserves identity and reverses composition : @xmath1631 , and @xmath1632 ( for @xmath1633 , @xmath1634 , and @xmath1635 ) . to make things more explicit ,
let @xmath1636 denote the category of sets , and let @xmath1637 denote the set of @xmath34-structures on @xmath1 . then for @xmath1633 , and fixed @xmath69
, we get a function @xmath1638 ; and the assignment @xmath1630 yields a functor @xmath1639 ( here @xmath1640 means opposite category ) . on the other hand , for fixed @xmath34 , and @xmath289 , recall that we have the classwise pullback operation @xmath1641 ; the assignment @xmath1642 is also ( contravariantly ) functorial , and furthermore @xmath1643 commutes with @xmath1644 .
so we in fact get a bifunctor @xmath1645 ( here @xmath509 denotes the product category ) , where for @xmath289 and @xmath1633 , @xmath1646 is the function @xmath1647 .
the following proposition says that the bifunctor @xmath1648 is `` representable in the second coordinate '' : for every @xmath1 , there is a canonical theory , namely @xmath385 ( the `` scott sentence '' of @xmath1 ) , such that @xmath34-structures on @xmath1 are in natural bijection with interpretations @xmath1649 .
[ thm : sigmae - rep ]
let @xmath300 be a countable borel equivalence relation .
let @xmath385 be the `` scott sentence '' , and let @xmath386 be the canonical structure given by [ thm : sigmae ] . then for any other theory @xmath109 and structure @xmath1073
, there is a unique interpretation @xmath1650 such that @xmath1651 .
this is illustrated by the following diagram : @xmath1652{\#h } \drar[models][swap]{\#a } & \sigma_e \\ & \sigma \uar[dashed][swap]{\alpha } \end{tikzcd}\ ] ] we recall the definition of @xmath385 from the proof of [ thm : sigmae ] .
let @xmath1653 be the language of @xmath385 . recall that we regard @xmath2 as a subspace of @xmath228 , and that the structure @xmath1654 is given by @xmath1655 .
now in order to have @xmath1651 , we must have , for each @xmath104-ary @xmath111 and @xmath235 , that @xmath1656 this can be achieved by letting @xmath1657 be a quantifier - free @xmath1569-formula such that @xmath1658 ; such @xmath1572 exists , as in the remarks following the proof of [ thm : sigmae ] . since @xmath1572 is quantifier - free
, we have @xmath1659 , whence ( @xmath398 ) holds .
furthermore , @xmath275 is an interpretation @xmath1660 , since ( from the proof of [ thm : sigmae ] ) every countable model of @xmath385 is isomorphic to @xmath1661 for some @xmath235 , and for such models we have @xmath1662 . to check that @xmath275 is unique ,
suppose @xmath1663 is another interpretation such that @xmath1664 .
then @xmath1665 obeys ( @xmath398 ) , so for every @xmath111 and @xmath235 , we have @xmath1666 . since every countable model of @xmath385 is isomorphic to some @xmath1661 , this means that @xmath1667 for every @xmath111 , i.e. , that @xmath1668 as interpretations @xmath1660 . as usual with representable bifunctors
, we may now extend @xmath1669 to a functor @xmath1670 where @xmath1671 , and for a class - bijective homomorphism @xmath289 , @xmath1672 is the unique interpretation ( given by [ thm : sigmae - rep ] ) such that @xmath1673 , as illustrated by the following diagram : @xmath1674[swap]{f } \rar[models]{\#h_e } \drar[models][pos=.35]{f^{-1}_e(\#h_f ) } & [ 2em ] \sigma_e \\[2em ] f \rar[models][swap]{\#h_f } & \sigma_f \uar[dashed][swap]{s(f ) } \end{tikzcd}\ ] ] then [ thm : sigmae ] says precisely that this functor is full and faithful , whence the functor @xmath24 is a contravariant equivalence of categories between @xmath248 and a full subcategory of @xmath1593 . for @xmath289 , unravelling the definitions and proofs reveals that @xmath1672 maps each @xmath8 in the language of @xmath477 coding the @xmath353th subbasic clopen subset of @xmath228 to a quantifier - free formula coding the preimage of that subset under @xmath47 .
it would be interesting to characterize the essential image of @xmath24 , i.e. , those theories which are isomorphic ( in the sense of @xmath1675 ) to @xmath385 for some countable borel equivalence relation @xmath1 .
we note that the operation @xmath1676 , being characterized by a universal property , can now be extended to a bifunctor @xmath1677 ( satisfying the obvious compatibility conditions ) .
finally , we note that the notion of interpretation gives us a convenient way of restating [ thm : einftys - smooth ] .
let @xmath1678 be a sentence in the language @xmath1679 , @xmath23 unary , asserting that @xmath23 defines a finite nonempty subset .
clearly @xmath1678 axiomatizes the smooth countable borel equivalence relations . for another theory @xmath109 ,
an interpretation @xmath1680 is the same thing as a choice of an @xmath33-formula @xmath1681 such that @xmath34 logically implies @xmath1682 , i.e. , @xmath34 logically implies that @xmath1572 defines a finite nonempty subset .
thus ( i)@xmath1367(ii ) in [ thm : einftys - smooth ] can be restated simply as @xmath1683 similarly , [ prb : elem - interp ] above asks whether for any @xmath816 , there is a @xmath1684 such that for any @xmath34 , @xmath1685 iff @xmath1686 .
there is a more abstract viewpoint on theories and interpretations , explaining much of the categorical structure in @xmath1593 , which we now sketch .
roughly , a theory @xmath109 can be seen as a `` presentation '' of a more intrinsic algebraic structure , where @xmath10 is the set of generators and @xmath34 is the relations ( combined into one ) .
an interpretation between theories is then simply a homomorphism of algebras ; and general universal - algebraic constructions may be used to combine theories .
the notion of `` @xmath1687-boolean algebra '' we are about to introduce is a special case of a much more general notion , _ hyperdoctrines _ , in categorical logic ( see @xcite ) .
similar notions have also appeared in model theory ( often in dualized forms ; see ( * ? ? ?
* section 3 ) , ( * ? ? ?
* section 2.2 ) ) .
however , for the sake of being self - contained , and because our point of view is different from both categorical logic and traditional model theory , we will not use these preexisting terminologies . let @xmath1688 denote the category of natural numbers and functions between them ( where as usual , @xmath1432 for @xmath253 ) .
recall that a boolean @xmath34-algebra is an @xmath83-complete boolean algebra ; to avoid confusion with sentences @xmath34 , we will use the term @xmath83-boolean algebra throughout this section .
let @xmath1689 denote the category of @xmath83-boolean algebras and @xmath83-homomorphisms ( i.e. , boolean algebra homomorphisms preserving countable joins ) .
an @xmath1687-boolean algebra is a functor @xmath1690 satisfying certain conditions , as follows . for a morphism @xmath1691 in @xmath1688 ,
the @xmath83-homomorphism @xmath1692 is required to have a left adjoint , denoted @xmath1693 , i.e. , a function ( not necessarily an @xmath83-homomorphism ) satisfying @xmath1694 for all @xmath1695 and @xmath1696 ( such @xmath1697 is uniquely determined by @xmath1698 ) .
the left adjoints are required to satisfy the beck - chevalley condition : for any pushout square @xmath1699{f_1 } \rar{f_2 } & m_2 \dar{g_2 } \\
m_1 \rar[swap]{g_1 } & n \end{tikzcd}\ ] ] in @xmath1688 ( i.e. , the square commutes , and exhibits @xmath104 as ( an isomorphic copy of ) the quotient of @xmath1700 by the smallest equivalence relation identifying @xmath1701 with @xmath1702 for each @xmath1703 ) , we have @xmath1704 as functions @xmath1705 .
( note : the @xmath1706 inequality is automatic . ) in fact , ( @xmath399 ) is symmetric ( under exchanging @xmath369 and @xmath219 in subscripts ) ; this can be seen by verifying that it is equivalent to : @xmath1707 for all @xmath1708 and @xmath1709 ( again , the @xmath1706 inequality is automatic ) .
a homomorphism between @xmath1687-boolean algebras @xmath1710 is a natural transformation of functors @xmath1711 , i.e. , a system of @xmath83-boolean homomorphisms @xmath1712 intertwining @xmath1713 for @xmath1691 in @xmath1688 , which furthermore commutes with the operations @xmath1697 , i.e. , @xmath1714 for all @xmath1691 in @xmath1688 , where @xmath1715 ( respectively @xmath1716 ) denotes the operation @xmath1697 in @xmath329 ( respectively @xmath1522 ) .
let @xmath1717 denote the category of @xmath1687-boolean algebras and homomorphisms . from its definition ,
the notion of @xmath1687-boolean algebra is clearly ( multisorted ) algebraic , in the sense that an @xmath1687-boolean algebra is a system of sets @xmath1718 equipped with various ( infinitary ) operations between them which are required to satisfy some universal equational axioms .
namely , the operations are : the boolean operations @xmath1719 , ( countable ) @xmath1720 on each @xmath1721 , and the unary operations @xmath1692 and @xmath1693 for each morphism @xmath1691 in @xmath1688 ; and the equational axioms are : the axioms of an @xmath83-boolean algebra for each @xmath1721 , the statements that each @xmath1698 is an @xmath83-homomorphism , and the axioms ( @xmath398 ) and ( @xmath399 ) for each choice of parameters @xmath1722 ( where ( @xmath398 ) is replaced with equivalent equations using @xmath1723 , say ) . furthermore
, a homomorphism of @xmath1687-boolean algebras is precisely a system of functions which are required to preserve all of the operations .
it follows from universal algebra that the category @xmath1717 admits all sorts of constructions .
we describe some of these , while at the same time relating @xmath1687-boolean algebras to theories .
let @xmath10 be a language .
for @xmath253 , let @xmath1724 denote the @xmath104-ary relation symbols in @xmath10 ; we may thus regard @xmath10 as an @xmath325-graded set @xmath1725 .
similarly , let @xmath1726 denote the @xmath33-formulas with free variables from @xmath1727 , and let @xmath1728 denote the logical equivalence classes of such formulas .
we identify each relation @xmath1729 with the atomic formula @xmath1730 .
each @xmath1731 is an @xmath83-boolean algebra under the logical connectives ( the lindenbaum - tarski algebra with @xmath104 free variables ) . for @xmath1691 in @xmath1688 ,
let @xmath1732 be the variable substitution map .
then @xmath1733 is the free @xmath1687-boolean algebra generated by @xmath10 , i.e. , by the atomic formulas @xmath1734 for @xmath1729 .
one verifies that a complete proof system for @xmath33 , e.g. , the gentzen system in @xcite , corresponds to the axioms of @xmath1687-boolean algebras .
the details are tedious but straightforward ; we only comment here on the treatment of quantifiers and equality in @xmath1687-boolean algebras . both
are encoded into the left adjoints @xmath1697 in an @xmath1687-boolean algebra .
namely , when @xmath1691 is an injective morphism in @xmath1688 , say ( for notational simplicity ) @xmath1735 is the inclusion , then @xmath1697 corresponds to existential quantification over the free variables not in the image of @xmath47 : @xmath1736 the axiom ( @xmath399 ) above ( with @xmath1737 ) then says that substitution of variables along @xmath1738 commutes with the quantifier @xmath1739 , while the axiom ( @xmath398 ) corresponds to the logical rule of inference characterizing the existential quantifier : @xmath1740 when @xmath1691 is surjective , say @xmath1741 is the identity on @xmath104 and @xmath1742 , then @xmath1697 corresponds to equating the variables in the kernel of @xmath47 : @xmath1743 again ( @xmath399 ) says that substitution preserves equality , while ( @xmath398 ) is the leibniz rule for equality .
we leave the details to the reader .
similarly , one can verify that for an @xmath33-sentence @xmath34 , @xmath1744 is the free @xmath1687-boolean algebra generated by @xmath10 , subject to the relation @xmath1745 ( where @xmath1595 is a tautology ) . but these are precisely the countably presented @xmath1687-boolean algebras , since countably many relations may be conjuncted into one ; hence the functor @xmath1746 ( sending interpretations to homomorphisms ) is an equivalence between the category @xmath1593 of theories and the full subcategory of @xmath1717 consisting of the countably presented @xmath1687-boolean algebras .
the category @xmath1717 , being ( multisorted ) algebraic , has all ( small ) limits and colimits .
we have already seen ( [ sec : interp - plus - times ] ) that the subcategory of countably presented algebras , or equivalently @xmath1747 , has countable products and coproducts .
the latter can be seen as an instance of the standard fact that countable colimits of countably presented algebras are countably presented ( see e.g. , ( * ? ? ? * vi 2.1 ) for the analogous statement for finitely presented algebras ) : the category @xmath1593 has all colimits of countable diagrams . for example , to construct the colimit of the diagram consisting of two theories @xmath109 and @xmath1566 and two interpretations @xmath1633 and @xmath1748 ( the coequalizer of @xmath1749 ) , one takes the `` quotient '' of @xmath1566 by the `` relations '' @xmath1750 for each @xmath111 , i.e. , the theory @xmath1751
recall the definition of @xmath83-distributive lattice from [ sec : lattice ] .
a boolean @xmath34-algebra is an @xmath83-complete boolean algebra , which is then automatically @xmath83-distributive .
the loomis - sikorski representation theorem states that every boolean @xmath34-algebra is a quotient of a sub - boolean @xmath34-algebra of @xmath913 for some set @xmath2 .
one proof ( see @xcite ) is via the stone representation theorem . by replacing the stone representation theorem with the priestley representation theorem for distributive lattices , we may prove an analogous result for @xmath83-distributive lattices .
we could not find this result in the literature , although it is quite possibly folklore .
we give the proof here for the sake of completeness .
[ thm : sdlat ] let @xmath776 be an @xmath83-distributive lattice
. then @xmath776 is isomorphic to a quotient @xmath83-complete lattice of a sub-@xmath83-complete lattice of @xmath913 for some set @xmath2 .
we recall the priestley representation theorem ; see ( * ? ? ?
* ii 4.58 , vii 1.1 ) .
let @xmath776 be a distributive lattice .
a prime filter on @xmath776 is an upward - closed subset @xmath1752 which is closed under finite meets ( including the top element ) and whose complement is closed under finite joins .
let @xmath2 be the set of prime filters on @xmath776 .
we order @xmath2 by @xmath75 . for @xmath1753 ,
let @xmath1754 we equip @xmath2 with the topology generated by the sets @xmath1755 and their complements . 1 .
@xmath2 is a compact hausdorff zero - dimensional space , and the order @xmath75 on @xmath2 is closed as a subset of @xmath132 .
the map @xmath1756 is an order - isomorphism between @xmath776 and the poset of clopen ( @xmath75-)upward - closed subsets of @xmath2 under inclusion .
now let @xmath776 be an @xmath83-distributive lattice , @xmath2 be as above , @xmath1757 be the @xmath34-ideal of meager subsets of @xmath2 , and @xmath1758 be the quotient map .
then @xmath1759 is a lattice homomorphism .
it is injective , since if @xmath1760 with @xmath967 then @xmath1761 by the baire category theorem .
clearly @xmath1765 contains each @xmath1766 and is closed ; thus @xmath1767 .
suppose there were some @xmath1768 .
so @xmath1769 .
let @xmath1770 for each @xmath1771 , we have some @xmath1772
. then @xmath1773 , so since @xmath601 is compact , there are @xmath1774 such that @xmath1775 . put @xmath1776 , so that @xmath1777
. then for all @xmath1778 , we have @xmath1779 , so by definition of @xmath601 , @xmath1780 .
also , we have @xmath1781 since @xmath43 is a filter .
now for each @xmath1782 , we find @xmath1783 ; then the union of @xmath1784 for all these @xmath240 covers @xmath1785 , so by compactness of the latter , by taking the join of finitely many @xmath1786 s , we get a @xmath1787 ( because @xmath43 is prime ) such that @xmath1788 .
but then @xmath1789 , i.e. , @xmath1790 , for every @xmath353 , while @xmath1791 because @xmath43 separates them , a contradiction . it follows from these lemmas that @xmath1759 is an @xmath83-complete lattice homomorphism ( i.e. , preserves countable meets and joins ) , so @xmath776 is isomorphic to its image @xmath1794 . but
@xmath1794 is a sub-@xmath83-complete lattice of a quotient of @xmath913 , hence ( as with any kind of algebra ) also a quotient of a sub-@xmath83-complete lattice of @xmath913 ( namely @xmath1795 ) .
[ thm : sdlat - thy ] if an equation between @xmath83-complete lattice terms ( i.e. , formal expressions built from variables , @xmath1796 , and @xmath1797 ) holds in the @xmath83-complete lattice @xmath219 , then it holds in every @xmath83-distributive lattice .
f. w. lawvere , equality in hyperdoctrines and comprehension schema as an adjoint functor , _ proc .
new york symposium on applications of categorical algebra _ ( ed .
a. heller ) , amer .
, 1970 , 114 . | for a class @xmath0 of countable relational structures , a countable borel equivalence relation @xmath1 is said to be @xmath0-structurable if there is a borel way to put a structure in @xmath0 on each @xmath1-equivalence class .
we study in this paper the global structure of the classes of @xmath0-structurable equivalence relations for various @xmath0 .
we show that @xmath0-structurability interacts well with several kinds of borel homomorphisms and reductions commonly used in the classification of countable borel equivalence relations .
we consider the poset of classes of @xmath0-structurable equivalence relations for various @xmath0 , under inclusion , and show that it is a distributive lattice ; this implies that the borel reducibility preordering among countable borel equivalence relations contains a large sublattice .
finally , we consider the effect on @xmath0-structurability of various model - theoretic properties of @xmath0 . in particular , we characterize the @xmath0 such that every @xmath0-structurable equivalence relation is smooth , answering a question of marks .
= 1 |
in the standard model of electroweak interactions all masses are generated by the higgs mechanism . as a consequence , at high temperatures a transition occurs from a massive low - temperature phase to a ` massless ' high - temperature phase , where the higgs vacuum expectation value ` evaporates ' and the electroweak symmetry is ` restored ' @xcite . due to the chiral nature of the weak interactions baryon number ( @xmath7 ) and lepton number ( @xmath8 ) are not conserved in the standard model @xcite . at zero temperature
this has no observable effect due to the smallness of the weak coupling .
however , as the temperature approaches the critical temperature @xmath9 of the electroweak phase transition , @xmath7 and @xmath8 violating processes come into thermal equilibrium @xcite .
their rate is determined by the free energy of sphaleron - type field configurations which carry topological charge . in the standard model
they induce an effective interaction of all left - handed fermions ( cf .
1 ) which violates baryon and lepton number by three units , b = l = 3 .
since @xmath7 and @xmath8 violating processes fall out of thermal equilibrium below @xmath9 , the presently observed value of the baryon asymmetry of the universe has finally been determined at the electroweak transition .
hence , the study of the thermodynamics of this transition is of great cosmological significance .
sphaleron processes conserve @xmath1 . in the high - temperature , ` symmetric ' phase , where @xmath7 and @xmath8 violating processes are expected to stay in thermal equilibrium over a wide range of temperatures , the expectation values of @xmath7 and @xmath8 are related , [ bleq ] b _ t c b - l _
t l _ t , where @xmath10 in the standard model @xcite .
hence , baryon and lepton number are correlated in the symmetric phase , and the generation of a baryon asymmetry requires lepton number violation . in the abelian higgs model , i.e. , scalar electrodynamics , the high - temperature phase transition is known to be of first order for sufficiently small higgs masses @xcite .
the expectation value of the higgs field plays the role of an order parameter . at temperatures just above the critical temperature @xmath9
the higgs phase is metastable .
it decays in a first - order transition to the symmetric phase , which is accompanied by a jump in the order parameter @xmath11 ( cf .
fig . 2 ) .
a similar behaviour is expected in the standard model , with a critical temperature given by the fermi constant , @xmath12 . in a first - order phase transition
a strong deviation from thermal equilibrium can occur .
since the standard model contains @xmath13 violating and , at high temperatures , also @xmath7 violating interactions , all conditions for baryogenesis are fulfilled .
it is therefore conceivable that the cosmological baryon asymmetry has indeed been generated at the electroweak phase transition @xcite .
such a scenario requires that a produced baryon asymmetry is not erased by the sphaleron processes in the higgs phase close to the critical temperature . from this condition
a lower bound on the jump of order parameter at the phase transition can be derived ( see @xcite ) , [ spcon ] > 1.2 .
this condition is necessary for electroweak baryogenesis although far from sufficient .
a discussion of the complicated nonequilibrium processes in the electroweak plasma can be found in @xcite . in order to examine whether condition ( [ spcon ] ) is satisfied
, one has to study the thermodynamics of the electroweak transition near the critical temperature . here
the main obstacle are the well known infrared divergencies of non - abelian gauge theories at high temperature . as a first step towards a treatment of the full standard model
the following simplifications are usually made : fermions are integrated out using high - temperature perturbation theory , and the electromagnetic interaction is neglected ( @xmath14 ) .
one is then left with the su(2 ) higgs model .
the observables which characterize a first - order phase transition are critical temperature , jump in the order parameter and latent heat . also important are surface tension and correlation lengths which are more difficult to compute and which we shall not discuss in detail .
our analysis will be based on the gauge invariant ` order parameter ' @xmath15 and the corresponding free energy computed in lattice simulations and perturbation theory . in our discussion
we shall closely follow @xcite .
it is well known that the electroweak phase transition is influenced by non - perturbative effects whose size is governed by the confinement scale of the effective three - dimensional theory which describes the high - temperature limit of the su(2 ) higgs model .
these effects are particularly relevant in the symmetric phase , and one may worry to what extent a perturbative analysis of the phase transition can yield sensible results .
as we shall see , perturbation theory is self - consistent at two - loop order .
a comparison with results obtained by lattice simulations will then enable us to estimate the effect of non - perturbative corrections .
+ _ generalities _ + the action of the su(2 ) higgs model at finite temperature @xmath16 reads [ st ] s_[,w ] = _ dx , with & = & ( + i ) , d_= ( _ - i g w _ ) , w _
= , + & & _ dx = _ 0^d_d^3x , = 1 t . here
@xmath17 is the vector field , @xmath18 is the higgs field , @xmath19 is the goldstone field , @xmath20 is the triplet of pauli matrices , and @xmath21 is the spatial volume
. for perturbative calculations gauge fixing and ghost terms have to be added to the action ( [ st ] )
. the free energy density of the system , @xmath22 , is given by the partition function , i.e. , the trace of the density matrix , ( -w(t , j ) ) = , where @xmath23 is the hamilton operator of the theory , and @xmath24 is the operator describing the higgs field . we have added a source @xmath25 , with @xmath26 , coupled to the spatial average of the gauge invariant composite operator @xmath27 ( here and below the trace operator acting on @xmath27 is omitted for brevity ) .
the partition function can be expressed as a euclidian functional integral ( see @xcite ) , [ fi ] ( -w(t , j ) ) = _ dd^ dw _
( -_dx(l + j ^ ) ) , where _ l _ is the euclidean lagrangian density , and the bosonic fields @xmath28 and @xmath29 satisfy periodic boundary conditions at @xmath30 and @xmath31 .
( [ fi ] ) is the starting point of perturbative as well as numerical evaluations of the free energy .
note , that the source @xmath25 in eq .
( [ fi ] ) couples to a gauge invariant composite field .
hence , the free energy @xmath22 is gauge independent .
the spatially constant source @xmath25 simply redefines the mass term in the action ( [ st ] ) .
this is in contrast to the usually considered generating function of connected green functions at zero momentum , ( -(t , j;j ) ) = _ dd^ dw_(-_dx(l + j ^+ j ) ) . here
the source @xmath32 couples to a gauge dependent quantity , the field @xmath18 . consequently , @xmath33 is gauge dependent and not a physical observable .
we have also kept the dependence on the source @xmath25 . from @xmath33
one obtains the effective potential @xmath34 via legendre transformation , with @xmath35 .
the wanted free energy density @xmath22 can now be obtained from the effective potential @xmath36 . in the infinite volume limit ,
one has [ conn ] w(t , j ) = ( t,_min(t , j),j ) , where @xmath37 is the global minimum of the effective potential @xmath38 . for arbitrary values of @xmath39 the potential @xmath36 is gauge dependent .
however , its value at the minimum is known to be gauge independent , yielding a gauge independent free energy @xmath22 . at the critical temperature @xmath9 of a first - order transition the order parameter @xmath40 , _ d^3x
^(x)(x ) = jw(t , j ) , and the energy density @xmath41 , e(t , j ) = w(t , j ) - t tw(t , j ) , are discontinuous .
the jump in the energy density is the latent heat @xmath42 .
for the first - order phase transition from liquid to vapour there exists a well known relation between the latent heat and the change of the molar volume , the clausius - clapeyron equation @xcite .
in the electroweak phase transition the order parameter @xmath43 plays the role of the molar volume , and a completely analogous relation can be derived .
the electroweak plasma can exist in two phases , the massive low - temperature higgs phase with free energy @xmath44 and the massless high - temperature symmetric phase with free energy @xmath45 . in the @xmath46-plane
the boundary between the two phases is determined by the equilibrium condition [ equi ] w_s(t ,
j(t ) ) = w_b(t , j(t ) ) . this equilibrium condition yields a useful connection between the latent heat @xmath47 and the jump in the order parameter @xmath48 . using the definitions q = - t t(w_s - w_b ) , = 2 j(w_s - w_b ) , one easily obtains [ cceq ] q = td jd t .
this is the clausius - clapeyron equation of the electroweak phase transition .
together with a dimensional analysis it implies q = - m_h^2 ( 1 + o(g^2 , ) ) , [ cc]where @xmath49 is the physical higgs mass at zero temperature .
this relation provides a useful check for perturbative as well as lattice results .
+ _ perturbation theory _ + near the ground state , @xmath50 , the free energy @xmath22 can be evaluated as power series in the couplings @xmath51 and @xmath52 by means of resummed perturbation theory which has been carried out up to two loops @xcite . here
, thermal corrections are added to the tree - level masses of the scalar fields and the longitudinal component of the vector boson field , [ ct ] s _
= d^3x ( _ 01 t^2 ( ^2 + ^2 ) + _ 1 t^2 w_l^2 ) . the sum of tree - level masses and thermal corrections then enters the boson propagators in loop diagrams , and @xmath53 is treated as counter term .
it turns out that the resummation of static modes only is a preferred procedure @xcite . hence , in eq .
( [ ct ] ) the fields @xmath18 and @xmath54 do not depend on the imaginary time @xmath55 . to leading order in the couplings ,
one obtains for the parameters in eq .
( [ ct ] ) from one - loop self energy corrections @xmath56 , @xmath57 .
the masses of the boson propagators are obtained from eqs .
( [ st ] ) and ( [ ct ] ) by shifting the higgs field @xmath18 by the average field @xmath39 .
this yields @xmath58 , @xmath59 , @xmath60 and @xmath61 for longitudinal and transverse part of the vector field , the higgs field and the goldstone field , respectively .
the resummation procedure can be optimized by adding terms of higher order in the couplings to eq .
( [ ct ] ) . in the higgs phase , where one is only interested in the effective potential close to the minimum
, the choice of a field dependent correction @xmath62 is useful which yields for the scalar masses , [ msb ] m_^2 = 2 ^ 2 , m_^2 = 0 .
hence , no thermal resummation for scalar masses is performed in the higgs phase . in the symmetric phase it is useful to determine the scalar masses self - consistently by [ selfc ] m_^2 = m_^2 = 1(t,0;j ) . for given vector boson masses @xmath58 and @xmath59 , this is a gap equation for the scalar masses , which can be solved at each order of the loop expansion .
having specified the vector boson and scalar masses in the higgs phase and in the symmetric phase , the free energy can be calculated from the effective potential @xmath36 , using the relation ( [ conn ] ) .
its two local minima yield the free energy @xmath45 and @xmath44 in the symmetric phase and the higgs phase , respectively .
the free energy of the ground state is w(t,0 ) = \{w_s(t,0 ) , w_b(t,0 ) } .
it is a concave function whose derivative is discontinuous at the critical temperature @xmath9 .
we can also consider the dependence of the free energy on the external source at the critical temperature @xmath9 .
this function , @xmath63 , is shown in fig .
we have subtracted in both phases the huge linear term @xmath64 , whose sole effect is to shift the expectation value @xmath15 by @xmath65 . from the free energy @xmath22
one can obtain the gauge invariant effective potential @xmath66 by means of a legendre transformation . since the derivative of @xmath22 is not continuous everywhere
, one has to use the definition ( see @xcite ) , v(t , ) = _ j\{w(t , j ) - j } .
this yields the convex , non - analytic function , which is plotted in fig . 4 as full line .
one may also compute the ordinary legendre transform @xmath67 and @xmath68 of @xmath45 and @xmath44 , respectively . @xmath69 and @xmath70 are also shown in fig .
4 . in the region outside of the two local minima , @xmath69 and @xmath70 , respectively , coincide with the convex effective potential @xmath66 .
between the two local minima , @xmath69 and @xmath70 represent two analytical continuations of @xmath66 , which meet at the ` matching point ' @xmath71 . at this point , marked by a cross in the plot ,
the first derivatives of both curves coincide .
the non - convex ` effective potential ' obtained by combining @xmath69 and @xmath70 on both sides of the ` matching point ' has a barrier between symmetric and higgs phase like the ordinary effective potential @xmath38 .
note , however , the difference in the range of fields . for @xmath36 one
has @xmath72 , whereas for the potential @xmath73 the field @xmath40 varies in the range @xmath74 .
the generation of a barrier between two local minima as analytic continuation from a convex effective potential is reminiscent of the treatment of first - order phase transitions in condensed matter physics @xcite .
however , the precise physical meaning of the resulting non - convex ` effective potential ' still remains to be understood .
+ + the asymmetric resummation described above has been carried out up to two - loop order . in the higgs phase
the scalar masses are given by eq .
( [ msb ] ) , and in the symmetric phase they are self - consistently determined from eq .
( [ selfc ] ) .
note , that in the two - loop calculation the counter term to be inserted in the one - loop graph is @xmath75 , whereas the scalar mass determined from eq .
( [ selfc ] ) is of higher order in @xmath51 . in the symmetric phase the self - consistently determined scalar masses are infrared divergent . with @xmath76 ,
the two - loop potential yields a contribution which diverges logarithmically at @xmath77 , [ ldiv ] m^2 _ - 33 g^4128 ^ 2t^2 .
following @xcite one may regularize this divergence by means of a ` magnetic mass ' term . in eq .
( [ ldiv ] ) one substitutes @xmath78 . in the following numerical results will be given for @xmath79 , which is obtained by one - loop gap equations @xcite .
the results change only insignificantly if the parameter @xmath80 is varied between 0.3 and 3.0 .
in addition to the resummation procedure one has to choose a renormalization scheme .
a good choice is the @xmath81-scheme with @xmath82 , supplemented by finite counter terms @xmath83 and @xmath84 which account for the most important zero - temperature renormalization effects .
given the two - loop potential in the symmetric phase and in the higgs phase , one can numerically determine the critical temperature @xmath9 , where the two potentials at their respective local minima are degenerate .
differentiation with respect to temperature and the external source @xmath25 then yields latent heat and jump in the order parameter @xmath85 .
the result for @xmath86 is shown in fig .
5 , labelled ` new resummation ' . here
the zero - temperature standard model values @xmath87 80 gev and @xmath88 have been used . in the case of the ` old resummation '
@xcite @xmath89 corresponds to the position of the second minimum .
the one - loop potential is @xmath90 , the two - loop potential is @xmath91 . as fig .
5 illustrates , the ` new resummation ' procedure improves the convergence significantly .
the relative change of an observable from one - loop to two - loop may be characterized by @xmath92 . for the jump in the order parameter @xmath93
, @xmath94 increases from @xmath95 at @xmath96 gev to @xmath97 at @xmath98 gev . above @xmath99
gev the convergence deteriorates rapidly , and the perturbative calculation is no longer self - consistent .
observables of the four - dimensional su(2 ) higgs model at finite temperature can be directly computed by means of monte carlo simulations on lattices with spatial size @xmath100 and temporal size @xmath101 .
such finite temperature simulations at small values of @xmath52 were initiated in @xcite , and high statistics simulations were performed in @xcite .
for the two higgs masses
@xmath102 gev and @xmath103 gev the quantities @xmath9 , @xmath47 and @xmath104 were computed on @xmath105 and @xmath106 lattices , and for @xmath107 gev a detailed study of the first - order phase transition was carried out on lattices of temporal size up to @xmath108 . in a complementary approach
detailed studies have also been carried out based on dimensional reduction where non - zero matsubara frequencies are first integrated out perturbatively .
recent results on dimensional reduction are described in @xcite where also references to previous work can be found .
numerical simulations were performed for the effective three dimensional theory which led to a quantitative description of the first - order phase transition for higgs masses up to @xmath109 gev @xcite .
= 11.0 cm = 9.57 cm in fig . 6 results for the jump in the order parameter , @xmath110 , are compared @xcite , which were obtained by simulations of the four - dimensional theory @xcite , the three - dimensional theory @xcite and perturbation theory @xcite , respectively .
the comparison is made for the parameter values @xmath111 gev and @xmath112 .
the statistical errors of the numerical simulations are so small that they are invisible in the figure .
the agreement between the three approaches is remarkable and certainly better than the systematic uncertainties of perturbation theory ( cf .
fig . 5 ) .
the continuum limit for the critical temperature has been studied in detail in @xcite for @xmath113 gev , i.e. , @xmath114 .
the results are shown in fig . 7 , where the error bars include the statistical error and an estimate of the systematic error .
a comparison is made with perturbation theory and the three - dimensional simulation .
the lattice simulations yield a critical temperature slightly below the result from perturbation theory , but the difference is not significant . the agreement between results from perturbation theory and non - perturbative lattice simulations is surprizing , since in the symmetric phase perturbation theory is known to be infrared divergent , which prevents a straightforward extension of the present two - loop calculation to three loops .
the infrared behaviour of a running gauge coupling has been studied in @xcite . in the symmetric phase the scalar masses depend logarithmically on an infrared cutoff of order the magnetic scale @xmath115 .
this is reminiscent of the debye screening length in pure gauge theory which , at two - loop order , also requires an infrared cutoff @xmath115 @xcite . here
, a non - perturbative definition of the debye screening length can be given such that the dominant contribution is given by the perturbative result and an additional non - perturbative contribution can be evaluated in a well - defined manner @xcite .
such a split into a dominant perturbative contribution and a non - perturbative remainder should also be possible for the free energy in the symmetric phase .
this would then justify the perturbative treatment of the first - order electroweak phase transition . for the pure gauge theory
this problem has been discussed in @xcite .
perturbation theory in the symmetric phase fails with respect to correlation lengths of gauge - invariant vector and scalar fields .
lattice simulations yield shorter correlation lengths for vector fields in the symmetric phase than in the higgs phase @xcite .
the contrary is true for the correlation length of the vector field evaluated in a fixed gauge @xcite - a puzzle which still remains to be resolved .
the latter screening lengths appear to be related to the critical higgs mass @xmath99 gev where the first - order phase transition turns into a smooth crossover @xcite .
independent of the intriguing non - perturbative features of the symmetric phase , which require further studies , it is now known that the necessary condition ( [ spcon ] ) for electroweak baryogenesis is not satisfied in the standard model . for higgs boson masses above the present experimental bound of 58 gev @xcite one has according to fig . 6 , < 0.7 , which is much smaller than the lower bound ( [ spcon ] ) . for appropriate choices of parameters it is possible to satisfy this bound in some extensions of the standard model , but in these models it has not been demonstrated that electroweak baryogenesis is indeed possible .
this strongly suggests that the baryon asymmetry has been generated in the high - temperature phase of the standard model , as it is the case for instance in conventional grand unified theories .
in the high temperature phase of the standard model the asymmetries of baryon number and lepton number are proportional in thermal equilibrium ( cf .
( [ bleq ] ) ) , b _ t c b - l _
t l _ t . in the standard model , as well as its unified extension based on the group su(5 ) ,
@xmath1 is conserved .
hence , no asymmetry in @xmath1 can be generated , and @xmath116 vanishes .
furthermore , as discussed above , baryogenesis at the electroweak phase transition appears unlikely . as a consequence ,
the non - vanishing of the baryon asymmetry is a strong argument for lepton number violation .
this is naturally realized by adding right - handed majorana neutrinos to the standard model .
this extension of the standard model can be embedded into grand unified theories with gauge groups containing so(10 ) @xcite .
heavy right - handed majorana neutrinos can also explain the smallness of the light neutrino masses via the see - saw mechanism @xcite .
the connection between baryon number and lepton number at high temperatures can be used to generate a baryon asymmetry .
this was suggested by fukugita and yanagida @xcite .
the primordial lepton asymmetry is generated by the out - of - equilibrium decay of heavy majorana neutrinos in the standard manner .
this mechanism has subsequently been studied by several authors @xcite , and it has been shown that the observed baryon asymmetry ( cf .
@xcite ) , y_b = n_bs=(0.6 - 1)10 ^ -10 , can be obtained for a wide range of parameters . in unified theories with right - handed neutrinos @xmath1 is in general spontaneously broken .
unification also restricts the new parameters which are introduced by adding right - handed neutrinos to the standard model . in so(10 ) unification it is natural to assume a similar pattern of mixings and masses for leptons and quarks .
this ansatz , together with the requirement of baryogenesis , also restricts the scale of @xmath1 breaking . the following discussion is closely related to @xcite . the most general lagrangian for couplings and masses of charged leptons and neutrinos is given by _ y = -g_le_r -g__r -12m_r + , where @xmath117 is the left - handed lepton doublet and @xmath118 is the standard model higgs doublet .
the vacuum expectation value of the higgs field @xmath119 generates dirac masses @xmath120 and @xmath121 for charged leptons and neutrinos , m_l = g_lv m_d = g_v , which are assumed to be much smaller than the majorana masses @xmath122 .
this yields light and heavy neutrinos k^_l+_l^c k , n_r+_r^c , with masses m_- k^m_d1mm_d^t k^ * , m_nm , [ seesaw ] as mass eigenstates . here
@xmath123 is a unitary matrix which relates weak and mass eigenstates .
since the heavy neutrinos @xmath124 are majorana fermions , their decay to lepton and higgs scalar violates lepton number . in the rest system
the decay width of @xmath124 reads at tree level , _ di:=_rs(n^i^+l)+ _
rs(n^i+)= m_i8(m_d^m_d)_iiv^2 .
[ decay ] interference between tree level and one - loop amplitudes ( cf .
8) yields the @xmath13 asymmetry @xcite & & _ i=18v^2(m_d^m_d)_ii_j f ( m_j^2m_i^2)[cpasymm ] + & & f(x)=.the corresponding maximal @xmath1 asymmetry is @xmath125 , where @xmath126 is the number of relativistic degrees of freedom ( cf .
@xcite ) . in order to generate the observed baryon asymmetry several conditions have to be fulfilled .
first , the @xmath13 asymmetry @xmath127 has to be large enough ; second , the out - of - equilibrium condition @xmath128 for the decaying heavy neutrino has to be fulfilled , where @xmath129 is the hubble parameter ; third , the @xmath8 violating interactions have to be sufficiently weak in order not to erase the generated lepton asymmetry .
these conditions tend to favour small masses for the light neutrinos and a large scale of @xmath1 breaking @xcite
. however , these constraints are not model independent .
the corresponding bounds on the light neutrino masses can be considerably relaxed if appropriate chiral symmetries are effectively conserved in some temperature range in the symmetric phase @xcite .
all these conditions are automatically taken into account if one integrates the boltzmann equations including all relevant interactions for the model under consideration .
the results discussed below are based on such an analysis using the boltzmann equations described in @xcite .
all three heavy neutrino families are taken into account as intermediate states whereas only the asymmetry generated by the lightest of the right - handed neutrinos is relevant , since the asymmetries generated by the heavier neutrinos are washed out .
+ _ neutrino masses and mixings _ + let us now consider a similar pattern of mixings and mass ratios for leptons and quarks , which is natural in so(10 ) unification .
such an ansatz is most transparent in a basis where all mass matrices are maximally diagonal .
in addition to real mass eigenvalues two mixing matrices appear .
one can always choose a basis for the lepton fields such that the mass matrices @xmath120 for the charged leptons and @xmath122 for the heavy majorana neutrinos @xmath124 are diagonal with real and positive eigenvalues , m_l= ( cccm_e&0&0 + 0&m_&0 + 0&0&m _ ) m= ( cccm_1&0&0 + 0&m_2&0 + 0&0&m_3 ) .
in this basis @xmath121 is a general complex matrix , which can be diagonalized by a biunitary transformation .
therefore , we can write @xmath121 in the form m_d = v ( ccc m_1&0&0 + 0&m_2&0 + 0&0&m_3 ) u^ , where @xmath73 and @xmath130 are unitary matrices and the @xmath131 are real and positive . in the absence of a majorana mass term @xmath73 and
@xmath130 would correspond to kobayashi - maskawa type mixing matrices of left- and right - handed charged currents , respectively . according to eqs .
( [ decay ] ) and ( [ cpasymm ] ) the @xmath13 asymmetry is determined by the mixings and phases present in the product @xmath132 , where the matrix @xmath73 drops out .
therefore , to leading order , the mixings and phases which are responsible for baryogenesis are entirely determined by the matrix @xmath130 .
correspondingly , the mixing matrix @xmath123 in the leptonic charged current , which determines @xmath13 violation and mixings of the light leptons , depends on mass ratios and mixing angles and phases of @xmath130 and @xmath73 .
hence , there is no direct connection between the @xmath13 violation and generation mixing at high and low energies .
consider now the mixing matrix @xmath130 .
one can factor out five phases , which yields u=^i^i_3^i_8u_1 ^i_3^i_8 , where the @xmath133 are the gell - mann matrices .
the remaining matrix @xmath134 depends on three mixing angles and one phase , like the kobayashi - maskawa matrix for quarks . in analogy to the quark mixing matrix we choose the wolfenstein parametrization @xcite as ansatz for @xmath134 , [ mm ] u_1= ( ccc 1-^22 & & a^3(-i ) + - & 1-^22 & a^2 + a^3(1 - -i ) & -a^2 & 1 ) , where @xmath135 and @xmath136 are of order one , while the mixing parameter @xmath137 is assumed to be small . for the masses @xmath131 and @xmath138 we assume a hierarchy like for up - type quarks , m_1=b^4m_3&m_2=c^2m_3&b , c=(1 ) + m_1=b^4m_3&m_2=c^2m_3&b , c=(1).[mmass ] for the eigenvalues @xmath131 of the dirac mass matrix this choice is motivated by so(10 ) unification .
the masses @xmath138 can not be degenerate , because in this case there exists a basis for @xmath139 such that @xmath140 , which implies that no baryon asymmetry is generated .
for simplicity the masses @xmath138 are assumed to scale like the dirac neutrino masses .
the light neutrino masses are given by the seesaw formula ( [ seesaw ] ) .
the matrix @xmath123 , which diagonalises the neutrino mass matrix , can be evaluated in powers of @xmath137 .
a straightforward calculation gives the following masses for the light neutrino mass eigenstates m__e&=&b^2|c+^4ib|^4 m__+(^6)[mne ] + m__&=&c^2|c+^4ib|bc ^2m__+(^4)[mnm ] + m__&=&m_3 ^ 2m_3+(^4).[mnt ] the @xmath13-asymmetry in the decay of the lightest right - handed neutrino @xmath141 is easily obtained from eqs .
( [ cpasymm ] ) and ( [ mm])-([mmass ] ) , _ 1=-^4 + ( ^6 ) .
this yields for the magnitude of the @xmath13 asymmetry , [ cpa ] |_1| ^4+(^6 ) .
how close the value of @xmath142 is to this upper bound depends on the phases @xmath143 , @xmath144 and @xmath145 . since @xmath146 , one can already conclude that a large value of the yukawa coupling @xmath147 will be preferred by this mechanism of baryogenesis .
this holds irrespective of the neutrino mixings .
+ _ numerical results [ results ] _ + to obtain a numerical value for the produced baryon asymmetry , one has to specify the free parameters in the ansatz ( [ mm])-([mmass ] ) . in the following
we will use as a constraint the value for the @xmath148-mass which is preferred by the msw explanation @xcite of the solar neutrino deficit ( cf .
@xcite ) , m__310 ^ -3 .
[ msw ] a generic choice for the free parameters is to take all @xmath149 parameters equal to one and to fix @xmath137 to a value which is of the same order as the @xmath137 parameter of the quark mixing matrix , a = b = c = b = c=|+i|1 , 0.1 .
[ p1 ] from eqs .
( [ mne])-([mnt ] ) , ( [ msw ] ) and ( [ p1 ] ) one now obtains , m__e810 ^ -6 , m__0.15 .[m1 ] finally , a second mass scale has to be specified . in unified theories based on so@xmath150
the dirac neutrino mass @xmath151 is naturally equal to the top - quark mass , [ 3 t ] m_3=m_t174 .
this determines the masses of the heavy majorana neutrinos @xmath124 , m_3 210 ^ 14 , [ m3 ] and , consequently , @xmath152 and @xmath153 . from eq .
( [ cpa ] ) one obtains the @xmath13 asymmetry @xmath154 , where we have assumed maximal phases .
the solution of the set of boltzmann equations discribed in @xcite now yields the @xmath1 asymmetry ( see fig .
9a ) , y_b - l 310 ^ -10 , which is indeed the correct order of magnitude .
the precise value depends on unknown phases .
( a ) ( b ) the large mass @xmath155 of the heavy majorana neutrino @xmath156 ( cf .
( [ m3 ] ) ) , suggests that @xmath1 is already broken at the unification scale @xmath157 gev , without any intermediate scale of symmetry breaking .
this large value of @xmath155 is a consequence of the choice @xmath158 . to test the sensitivity of the result for @xmath159 on this assumption , consider the alternative choice , @xmath160 gev , with all other parameters remaining unchanged . in this case one
obtains @xmath161 gev and @xmath162 for the mass of @xmath156 and the @xmath13 asymmetry , respectively . since the maximal @xmath1 asymmetry is @xmath163 ( cf .
@xcite ) , it is clear that the generated asymmetry will be too small .
the solutions of the boltzmann equations are shown in fig .
the generated asymmetry , @xmath164 , is too small by more than two orders of magnitude .
we conclude that high values for both masses @xmath151 and @xmath155 are preferred , which is natural in so(10 ) unification .
models for dark matter involving massive neutrinos favour a @xmath165-neutrino mass @xmath166 @xcite , which is significantly larger than the value given in ( [ m1 ] ) .
such a large value for the @xmath165-neutrino mass can be accomodated within the ansatz described in this section .
however , it does not correspond to the simplest choice of parameters and requires some fine - tuning . for
the mass of the heaviest majorana neutrino one obtains in this case @xmath167 gev . without an intermediate scale of symmetry breaking , the unification of gauge couplings appears to require low - energy supersymmetry .
this provides further sources for generating a @xmath1 asymmetry @xcite , whose size depends on additional assumptions . in this case ,
especially constraints on the reheating temperature @xcite and the possible role of preheating @xcite require further studies .
the observation that baryon and lepton number violating processes are in thermal equilibrium in the high - temperature phase of the standard model , is of crucial importance for the theory of baryogenesis .
in particular it implies that the presently observed cosmological baryon asymmetry has been finally determined at the electroweak phase transition . during the past three years a quantitative understanding of the first - order electroweak phase transition for higgs boson masses up to @xmath168 gev
has been achieved by means of analytical and numerical methods .
the transition to a crossover near @xmath169 and the full understanding of the high - temperature phase still require further work .
however , already now we know that for higgs boson masses above the lower bound obtained at lep , baryogenesis at the weak electroweak transition is very unlikely .
searching for alternatives to electroweak baryogenesis the connection between baryon number and lepton number in the symmetric phase is again crucial .
it allows to generate the baryon asymmetry from a lepton asymmetry , as suggested by fukugita and yanagida .
necessary ingredients are right - handed neutrinos and majorana masses , which appear naturally in so(10 ) unified theories .
the example described in the previous section demonstrates that this mechanism can explain the observed cosmological baryon asymmetry without any fine - tuning of parameters .
further sources for generating a lepton asymmetry exist in models with low - energy supersymmetry .
all this suggests that , hoping for further progress in theory and new experimental results on neutrino properties , we can look forward to an intriguing interplay between non - perturbative processes in the standard model , early universe cosmology and neutrino physics .
+ + + * acknowledgements * + + it is a pleasure to thank z. fodor , a. hebecker and m. plmacher for an enjoyable collaboration which led to the results described in this report .
i would also like to thank j. hein , k. jansen and i. montvay for their continuous help in understanding lattice results and for many valuable discussions .
last , but not least , i would like to thank the organizers of _
quarks 96 _ for a stimulating meeting and also for hospitality in yaroslavl .
99 d. a. kirzhnits , jetp lett . 15 ( 1972 )
529 ; + d. a. kirzhnits and a. d. linde , phys .
b42 ( 1972 ) 471 g. t hooft , v. a. kuzmin , v. a. rubakov and m. e. shaposhnikov , phys .
b155 + ( 1985 ) 36 s. yu . khlebnikov and m. e. shaposhnikov b. i. halperin , t. c. lubensky and s .- k .
ma , d. a. kirzhnits and a. d. linde , for a discussion and references , see + a. g. cohen , d. b. kaplan and a. e. nelson , annu .
nucl . part .
43 ( 1993 ) 27 ; + v. a. rubakov and m. e. shaposhnikov , usp . fis .
nauk . 166
( 1996 ) 493 w. buchmller , z. fodor and a. hebecker , nucl .
b447 ( 1995 ) 131 j.i .
kapusta , _ finite - temperature field theory _
( cambridge univ . press , 1989 ) h. b. callan , _
( wiley , new york , 1960 ) p. arnold and o. espinosa , phys .
d47 ( 1993 ) 3546 ; d50 ( 1994 ) 6662 ( e ) z. fodor and a. hebecker , nucl
b432 ( 1994 ) 127 l. oraifeartaigh , a. wipf and h. yoneyama , nucl .
b271 ( 1986 ) 653 j.s .
langer , physica a73 ( 1974 ) 61 w. buchmller , z. fodor , t. helbig and d. walliser , ann .
( 1994 ) 260 j. r. espinosa , m. quirs and f. zwirner , phys .
b314 ( 1993 ) 206 b. bunk , e .-
ilgenfritz , j. kripfganz and a. schiller , nucl .
b403 + ( 1993 ) 453 z. fodor , j. hein , k. jansen , a. jaster and i. montvay , f. csikor , z. fodor , j. hein , a. jaster and i. montvay , a. kerres , g. mack and g. palma , a. jakovc and a. patks , hep - ph/9609364 k. kajantie , m. laine , k. rummukainen and m. shaposhnikov , m. grtler , h. perlt , a. schiller , e .-
ilgenfritz and j. kripfganz , desy 96 - 113 , hep - lat/9605042 k. jansen , nucl .
b ( proc . suppl . ) 47 ( 1996 ) 196 m. laine , hd - thep-96 - 08 , hep - lat/9604011 m. reuter and c. wetterich , nucl .
b408 ( 1993 ) 91 a. k. rebhan , p. arnold and l. g. yaffe , e. braaten and a. nieto , o. philipsen , m. teper and h. wittig , w. buchmller and o. philipsen , f. karsch , t. neuhaus , a. patks and j. rank , k. kajantie , m. laine , k. rummukainen and m. shaposhnikov , k. rummukainen , hep - lat/9608079 , t. yanagida , in _ workshop on unified theories _
( kek report 79 - 18 , 1979 ) , eds . o. sawada and a. sugamoto , p.
95 + m. gell - mann et al .
, in _ supergravity _
( north holland , amsterdam , 1979 ) , eds . p. van nieuwenhuizen and d. freedman , p. 315
m. fukugita and t. yanagida , ; a. e. nelson and s. m. barr , ; j. harvey and m. s. turner , ; b. a. campbell , s. davidson , j. ellis and k. a. olive , ; w. fischler , g. f. giudice , r. g. leigh and s. paban ; w. buchmller and t. yanagida , h. murayama et al . , ; b. a. campbell , s. davidson and k. a. olive , ; l. covi , e. roulet and f. vissani , preprint ic/96/73 , hep - ph/9605319 ; h. murayama and t. yanagida , ; m. dine , l. randall and s. thomas , | the presently observed cosmological baryon asymmetry has been finally determined at the time of the electroweak phase transition , when baryon and lepton number violating interactions fell out of thermal equilibrium .
we discuss the thermodynamics of the phase transition based on the free energy of the su(2 ) higgs model at finite temperature , which has been studied in perturbation theory and lattice simulations .
the results suggest that the baryon asymmetry has been generated by lepton number violating interactions in the symmetric phase of the standard model , i.e. , at temperatures above the critical temperature of the electroweak transition .
the observed value of the baryon asymmetry , @xmath0 , is naturally obtained in an extension of the standard model with right - handed neutrinos where @xmath1 is broken at the unification scale @xmath2 gev . the corresponding pattern of masses and mixings of the light neutrinos @xmath3 , @xmath4 and @xmath5 is briefly described .
= = = = = = = = = = = = = = = = = = = ' '' '' = = = = = = = # 1 = = = = = = = = = = = = = = = = = = ps .
= = = = = = = = = = = # 1#1 # 1 # 1*#1 * # 1@xmath6 = = = = = = = = = = = = = # 1*#1 * #
1#1 # 1_#1 _ ii = = = c=== = = = ==u c i u # 1 # 1 # 1^#1 # 1_#1 # 1/#1 /#1#1 #
1#1 # 1#1 # 1#1 # 1 # 1 # 1 # 1#1| # 1| # 1 # 1#1 # 1| # 1| # 1 # 1 # 1 # 1 # 1 # 1#2#1#2 # 1#2#1#2 # 1#2#3 ^ 2 # 1#2 # 3 # 1#2^#1 # 1 # 1 # 1#2#3phys .
lett .
b#1 ( 19#2 ) # 3 # 1#2#3nucl .
phys .
b#1 ( 19#2 ) # 3 # 1#2#3phys .
rev .
lett . # 1 ( 19#2 )
# 3 # 1#2#3phys .
rev .
d#1 ( 19#2 ) # 3 # 1#2#3class . and quantum grav .
# 1 ( 19#2 ) # 3 # 1#2#3commun .
math .
phys . # 1 ( 19#2 )
# 3 # 1#2#3j . math .
phys . # 1 ( 19#2 )
# 3 # 1#2#3ann . of phys .
# 1 ( 19#2 ) # 3 # 1#2#3phys
. rep .
# 1c ( 19#2 ) # 3 # 1#2#3progr .
theor .
phys . # 1 ( 19#2 )
# 3 # 1#2#3int .
j. mod .
phys .
a#1 ( 19#2 ) # 3 # 1#2#3mod .
phys .
lett .
a#1 ( 19#2 ) # 3 # 1#2#3nuovo cim . # 1 ( 19#2 )
# 3 # 1#2#3_ibid .
_ # 1 ( 19#2 ) # 3 = 14pt = 6.0 in = 8.5 in -0.25truein 0.30truein 0.30truein |
our approach to understanding the evolution of genetic architectures combines standard models from quantitative genetics @xcite with the wright - fisher model from population genetics @xcite . in its simplest version ,
our model considers a continuous trait whose value , @xmath4 , is influenced by @xmath5 loci .
each locus @xmath6 contributes additively an amount @xmath7 , so that the trait value is defined as the mean of the @xmath7 values across contributing loci .
this trait definition means that a gene s contribution to a trait is diluted when @xmath5 is large , which prevents direct selection on gene copy numbers when genes have similar contributions @xcite .
we discuss this definition below , along with alternatives such as the sum .
the fitness of an individual with trait value @xmath4 is assumed gaussian with mean @xmath8 and standard deviation @xmath9 , so that smaller values of @xmath9 correspond to stronger stabilizing selection on the trait @xcite .
individuals in a population of size @xmath10 replicate according to their relative fitnesses . upon replication ,
an offspring may acquire a point mutation that alters the direct effect of one locus , @xmath6 , perturbing the value of @xmath7 for the offspring by a normal deviate ; or the offspring may experience a duplication or a deletion in a contributing locus , which changes the number of loci @xmath5 that control the trait value in that individual ( see methods ) .
point mutations , duplications , and deletions occur at rates @xmath11 , @xmath12 , @xmath13 , which have comparable magnitudes in nature ( table s1 ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
finally , an offspring may also increase the number of loci that contribute to its trait value by recruitment that is , by acquiring a recruitment mutation , with probability @xmath14 , in some gene that did not previously contribute to the trait value ( see methods ) . over successive generations in our model , the genetic architecture underlying the trait
that is , how many loci contribute to the trait s value , and the extent of their contributions
varies among the individuals in the population , and evolves .
the genetic architectures that evolve in our model represent the complete genetic determinants of a trait , which may include but do not correspond precisely to the genetic loci that would be detected based on polymorphisms segregating in a sample of individuals in a qtl study .
we discuss this important distinction below , when we compare the predictions of our model to empirical qtl data .
we studied the evolution of genetic architectures in sets of @xmath15 replicate populations , simulated by monte carlo , with different amounts of selection on the trait .
we ran each of these simulations for @xmath16 million generations , in order to model the extensive evolutionary divergence over which genetic architectures are assembled in nature .
the form of the genetic architecture that evolves in our model depends critically on the strength of selection on the trait .
in particular , we found a striking non - monotonic pattern : the equilibrium number of loci that influence a trait is greatest when the strength of selection on the trait is intermediate ( fig .
1 ) . moreover , the variability in the contributions of loci to the trait value ( fig .
s1 ) and the effects of deleting or duplicating genes ( fig .
s2 ) are also greatest for a trait under intermediate selection . in other words ,
our model predicts that traits under moderate selection will be encoded by many loci with highly divergent effects ; whereas traits under strong or weak selection will be encoded by relatively few loci .
we also studied how epistatic interactions among loci influence the evolution of genetic architecture . to incorporate the influence of locus @xmath17 on the contribution of locus @xmath6 we introduced epistasis parameters @xmath18 so that the trait value is now given by @xmath19 where @xmath20 is a standard sigmoidal filter function ( * ? ? ? * see methods and fig .
s4 ) . as with the direct effects of loci
, the epistatic effects were allowed to mutate and vary within the population , and evolve .
although significant epistatic interactions emerge in the evolved populations ( fig .
s3b ) , the presence of epistasis does not strongly affect the average number of loci that control a trait ( figs .
s3a and s4 ) .
epistasis is not required for the evolution of large @xmath5 , nor does it change the shape of its dependence on the strength of selection .
there is an intuitive explanation for the non - monotonic relationship between the selection pressure on a trait and the number of loci that control it . for a trait under weak selection ( high @xmath9 ) ,
changes in the trait value have little effect on fitness .
thus , even if deletions , recruitments and duplications change the trait value , these changes are nearly neutral ( fig . 2 ) .
as a result , the number of loci controlling the trait evolves to its neutral equilibrium , which is small because deletions are more frequent than duplications and recruitments ( see methods , figs . 1 and s3 ) . on the other hand , when selection on a trait is very strong ( low @xmath9 ) , few point mutations , and only those with small effects on the trait , will fix in the population . as a result ,
all loci have similar contributions to the trait value ( fig . 2 row 1 ) , and so duplications or deletions again have little effect on the trait or on fitness ( fig .
2 rows 2 and 3 ) . in this case , the equilibrium number of loci is given by the value expected when deletions and duplications , but not recruitments , are neutral ( figs . 1 and s3 ) .
only when selection on a trait is moderate can variation in the contributions across loci accrue and impact the fixation of deletions and duplications ( fig . 2 row 4 ) , by a process called compensation : a slightly deleterious point mutation at one locus , which perturbs the trait value , segregates long enough to be compensated by point mutations at other loci @xcite .
compensation increases the variance in the contributions among loci ( fig .
2 , row 1 ) , as has been observed for many phenotypes in plants and animals @xcite . finally , even though duplications and deletions are mildly deleterious in this regime , there is a bias favoring duplications over deletions ( fig . 2 row 3 ) .
this bias arises because duplications increase the number of loci in the architecture , which attenuates the effect of each locus on the trait ( fig . 2 row 2 ) .
thus when selection is moderate , duplications and recruitments fix more often than deletions and drive the number of contributing loci above its neutral expectation ( fig . 2 rows 4 and 5 ) .
as the number of loci increases the bias is reduced ( fig . 2 rows 4 and 5 ) , and so @xmath5 equilibrates at a predictable value ( figs . 1 and s3 ) .
duplications and recruitments might also be slightly favored over deletions under intermediate selection , because architectures with more loci also have reduced genetic variation @xcite .
this effect which would positively select for an increase in gene copy numbers is likely weak in our model , as duplications and recruitments are deleterious on average under intermediate selection , only less so than deletions ( fig . 2 rows 4 and 5 ) .
( line 1 ) , and we then measured the effects of recruitments , deletions and duplications on the trait value ( line 2 ) and on fitness ( line 3 ) . from the latter , we calculated the rate at which deletions , recruitment and duplications enter and fix in the population ( line 4 ) , and the resulting rate of change in the number of loci contributing to the trait ( line 5 ) .
* line 1 : * for @xmath21 , the variation in direct effects ( @xmath7 ) and indirect effects among controlling loci ( @xmath22 ) increases as selection on the trait is relaxed .
* line 2 : * as a consequence of this variation among loci , the average change in the trait value following a duplication or a deletion also increases as selection on the trait is relaxed . * line 3 : * changes in the trait value are not directly proportional to fitness costs , because the same change in @xmath4 has milder fitness consequences when selection is weaker ( larger @xmath9 ) . as a result , the average fitness detriment of duplications and deletions is highest for traits under intermediate selection .
* line 4 : * consequently , the fixation rates of duplications and deletions are smallest under intermediate selection . *
line 5 : * the equilibrium number of loci controlling a trait under a given strength of selection is determined by that value of @xmath5 for which duplications and recruitments on one side , and deletions on the other , enter and fix in the population at the same rate .
for example , when @xmath23 these rates are equal when @xmath5 is close to @xmath24 ( black arrow ) , so that the equilibrium genetic architecture contains @xmath25 loci on average ( compare fig .
s3 black arrow).,width=680 ] the predictions of our model notably , that the number of loci in a genetic architecture is greatest for traits under intermediate selection
are robust to choices of population - genetic parameters .
the non - monotonic relation between selection pressure on a trait and the size of its genetic architecture , @xmath5 , holds regardless of population size ; but the location of maximum @xmath5 is shifted towards weaker selection in larger populations ( fig .
this result is compatible with our explanation involving compensatory evolution : selection is more efficient in large populations , and so compensatory evolution occurs at smaller selection coefficients . likewise , when the mutation rate is smaller the resulting equilibrium number of controlling loci is reduced ( fig .
this result is again compatible with the explanation of compensatory evolution , which requires frequent mutations .
increasing the rate of deletions relative to duplications also reduces the equilibrium number of loci in the genetic architecture , but our qualitative results are not affected even when @xmath13 is twice as large as @xmath12 ( fig .
s7 ) . finally , increasing the rate of recruitment @xmath26 ( or the genome size ) increases the number of loci contributing to all traits except those under very strong selection , as expected from fig .
our prediction that traits under intermediate selection are encoded by the richest genetic architectures is insensitive to changes in this parameter , and it holds even in the absence of recruitment ( fig .
our analysis has relied on several quantitative - genetic assumptions , which can be relaxed .
first , we assumed that all effects of locus @xmath6 ( _ i.e. _ @xmath7 and all @xmath27 and @xmath18 ) are simultaneously perturbed by a point mutation .
relaxing this assumption , so that a subset of the effects are perturbed , does not change our results qualitatively ( fig .
second , we assumed that point mutations have unbounded effects so that variation across loci can increase indefinitely . to relax this assumption we made mutations less perturbative to loci with large effects ( see methods ) .
even a strong mutation bias of this type led to very small changes in the equilibrium behavior ( fig .
third , we assumed no metabolic cost of additional loci , even though additional genes in _ saccharomyces cerevisiae _ are known to decrease fitness slightly @xcite . nonetheless , including a metabolic cost proportional to @xmath5 does not alter our qualitative predictions ( fig .
s11 ) . finally , we defined the trait value as the average of the contributions @xmath7 across loci , as opposed to their sum .
this definition reflects the intuitive notion that a gene product s contribution to a trait will generally depend on its abundance relative to all other contributing gene products .
moreover , this assumption that increasing the number of loci influencing a trait attenuates the effect of each one is supported by empirical data : changing a gene s copy number is known to have milder phenotypic effects when the gene has many duplicates @xcite . nonetheless
, alternative definitions of the trait value , which span from the sum to the average of contributions across loci , generically exhibit the same qualitative results ( text s1 and fig .
although robust to model formulation and parameter values , our results do depend in part on initial conditions .
when selection is strong , the initial genetic architecture can affect the evolutionary dynamics of the number of loci ( fig .
this occurs because the initial architecture may set dependencies among loci that prevent a reduction of their number .
this result indicates that only those architectures of traits under very strong selection should depend on historical contingencies .
we have also studied a multitrait version of our model , where genes participating in other traits can be recruited or lost through mutation . even though this model features pleiotropy , and
the effects of recruitments evolve neutrally , our qualitative results remained unaffected ( text s3 and fig .
s15 ) .
previous models related to genetic architecture have been used to study the evolutionary fate of gene duplicates .
these models typically assume that a gene has several sub - functions , which can be gained ( neo - functionalization ; * ? ? ? * ) or lost ( sub - functionalization ; * ? ? ?
* ; * ? ? ?
* ) in one of two copies of a gene .
such `` fate - determining mutations '' @xcite stabilize the two copies , as they make subsequent deletions deleterious .
such models complement our approach , by providing insight into the evolution of discrete , as opposed to continuous or quantitative , phenotypes .
yet there are several qualitative differences between our analysis and previous studies of gene duplication .
most important , our model considers the dynamics of both duplications and deletions , in the presence of point mutations that perturb the contributions of loci to a trait .
this co - incidence of timescales is important in the light of empirical data @xcite showing that changes in copy numbers occur at similar rates as point mutations ( table s1 ) . under these circumstances ,
a gene may be deleted or acquire a loss - of - function mutation before a new function is gained or lost .
our model includes these realistic rates , and accordingly we find that duplicates are very rarely stabilized by subsequent point mutations . instead , the number of loci in a genetic architecture may increase , in our model , because compensatory point mutations introduce a bias towards the fixation of duplications as opposed to deletions . like most evolutionary models
, our analysis greatly simplifies the mechanistic details of how specific traits influence fitness in specific organisms . as a result
, our analysis explains only the broadest , qualitative features of how genetic architectures vary among phenotypic traits , leaving a large amount of variation unexplained .
this remaining variation may be partly random ( as predicted by the distributions of the number of evolving loci , see _
e.g. _ fig .
1 ) , and partly due to ecological and developmental details that our model neglects . due to this variation , a quantitative comparison between our model and empirical data would require information about the genetic architectures for at least hundreds of traits ( see below , for our analysis of expression qtls ) . nevertheless , the qualitative , non - monotonic predictions of our model ( fig .
1 ) may help to explain some well - known trends in the genetics of human traits .
for instance , in accordance with our predictions , human traits under moderate selection , such as stature or susceptibility to mid - life diseases like diabetes , cancer , or heart - disease , are typically complex and highly polygenic ; whereas traits under very strong selection , such as those ( _ e.g. _ mucus composition or blood clotting ) affected by childhood - lethal disease like cystic fibrosis or haemophilias are often mendelian ; and so too traits under very weak selection ( such as handedness , bitter taste , or hitchhiker s thumb ) are often mendelian .
our analysis provides an evolutionary explanation for these differences , and it delineates the selective conditions under which we may expect a mendelian , as opposed to fisherian , architecture .
we tested our evolutionary model of genetic architectures by comparison with empirical data on a large number of traits .
such a comparison must , of course , account for the fact that our model describes the true genetic architecture underlying a trait , whereas any qtl study has limited power and describes only the associations detected from polymorphisms segregrating in a particular sample of individuals .
accounting for this discrepancy ( see below ) , we compared our model to data from the study of @xcite , who measured mrna expression levels and genetic markers in @xmath28 recombinant strains produced from two divergent lines of _ s. cerevisiae_. for each yeast transcript we computed the number of non - contiguous markers associated with transcript level , at a given false discovery rate ( see methods ) .
we also calculated the codon adaptation index ( cai ) of each transcript an index that correlates with the gene s wildtype expression level and with its overall importance to cellular fitness @xcite .
we found a striking , non - monotonic relationship between the cai of a transcript and the number of loci linked to variation in its abundance ( fig .
thus , assuming that cai correlates with the strength of selection on a transcript , @xcite detected more loci regulating yeast transcripts under intermediate selection than transcripts under either strong or weak selection .
we compared the empirical data on yeast eqtls ( fig .
3a ) to the predictions of our evolutionary model . in order to make this comparison , we first evolved genetic architectures for traits under various amounts of selection ( fig .
s3 ) , and for each architecture we then simulated a qtl study of the exact same type and power as the yeast eqtl study : that is , we generated 112 crosses from two divergent lines using the yeast genetic map ( text s2 ) .
as expected , the simulated qtl studies based these 112 segregants detected many fewer loci linked to a trait than in fact contribute to the trait in the true , underlying genetic architecture ( fig .
3b versus fig . 1 ) .
this result is consistent with previous interpretations of empirical eqtl studies @xcite .
the simulated qtl studies revealed another important bias : a locus that contributes to a trait under weak selection is more likely to be correctly identified in a qtl study than a locus that contributes to a trait under strong selection ( fig .
furthermore , our simulations demonstrate that the number of associations detected in such a qtl study depends on the divergence time between the parental strains used to generate recombinant lines ( fig .
finally , traits under weaker selection may be more prone to measurement noise , which we also simulated ( fig .
s18 ) . despite these detection biases , which we have quantified , the relationship between the selection pressure on a trait and the number of _ detected _ qtls in our model ( fig .
3b and figs .
s18 and s19 ) agrees with the relationship observed in the yeast eqtl data ( fig .
importantly , both of these relationships exhibit the same qualitative trend : traits under intermediate selection are encoded by the richest genetic architectures .
populations of genetic architectures , using the parameters corresponding to fig .
s3 . from each such population
, we then evolved two lines independently for @xmath29 generations in the absence of deletions , duplications and recruitment , to mimic the divergent strains used in the yeast cross of @xcite .
from these two divergent genotypes we then created @xmath28 recombinant lines following the genetic map from @xcite .
we then analyzed the resulting simulated data with _
r / qtl _ in the same way as we had analyzed the yeast data ( text s2 ) .
the distribution of qtls detected and their means are represented as in fig .
1 , for each value of selection strength @xmath9.,width=680 ]
many interesting developments lie ahead .
our model is far too simple to account for tissue- and time - specific gene expression , dominance , context - dependent effects , etc @xcite . how these complexities will change predictions for the evolution of genetic architectures
remains an open question .
nonetheless , our analysis shows that it is possible to study the evolution of genetic architecture from first principles , to form _ a priori _ expectations for the architectures underlying different traits , and to reconcile these theories with the expanding body of qtl studies on molecular , cellular , and organismal phenotypes .
we described the evolution of genetic architectures using the wright - fisher model of a replicating population of size @xmath10 , in which haploid individuals are chosen to reproduce each generation according to their relative fitnesses .
the fitness of an individual with @xmath5 loci encoding trait value @xmath4 is @xmath30 where @xmath31 denotes the density at @xmath4 of a gaussian distribution with mean @xmath8 and standard deviation @xmath32 , and the second term denotes the metabolic cost of harboring @xmath5 loci , which depends on a parameter @xmath33 .
the trait value of such an individual , given the direct contributions @xmath7 and epistatic terms @xmath18 is described by eq .
( 1 ) where @xmath34 is a sigmoidal curve , so that the epistatic interactions either diminish or augment the direct contribution of locus @xmath6 depending on whether @xmath35 is positive or negative ( fig .
s4 ) . in general , loci do not influence themselves ( @xmath36 ) and , in the model without epistasis , all @xmath37 and @xmath38 .
if an individual chosen to reproduce experiences a duplication at locus @xmath6 then the new duplicate , labelled @xmath39 , inherits its direct effect ( @xmath40 ) and all interaction terms ( @xmath41 and @xmath42 for all @xmath43 ) , with the interaction terms @xmath44 and @xmath45 initially set to zero .
recruitment occurs with probability @xmath26 per mutation of one of the @xmath46 genes not contributing to the trait .
the initial direct contribution @xmath7 of recruited locus @xmath6 is drawn from a normal distribution with mean zero and standard deviation @xmath47 ; its interaction terms with other loci ( @xmath39 ) , @xmath44 and @xmath45 , are initially set to zero .
note that this assumption is relaxed in the multilocus version of our model , where the direct and indirect effects of recruitments evolve neutrally ( text s3 and fig .
s15 ) . in general a point mutation at locus @xmath6 changes its contribution to the trait , @xmath7 , and all its epistatic interactions , @xmath27 and @xmath18 , each by an independent amount drawn from a normal distribution with mean zero and standard deviation @xmath47 .
the normal distribution satisfies the assumptions that small mutations are more frequent than large ones @xcite , and that there is no mutation pressure on the trait @xcite .
we relaxed the former assumption by drawing mutational effects from a uniform distribution without qualitative changes to our results ( fig .
s13 ) . in order to relax the latter assumption we included a bias towards smaller mutations in loci with large effects , so that the mean effect of a mutation at locus @xmath6 now equals @xmath48 and @xmath49 , respectively for @xmath7 and @xmath27 @xcite .
we also considered a model in which a mutation at locus @xmath6 affects only a proportion @xmath50 of the values @xmath7 , @xmath27 , and @xmath18 . by default
, simulations were initialized with @xmath51 and @xmath52 ; alternative initial conditions were also studied , as shown in fig .
s14 . when deletions and duplications are neutral , and recruitments strongly deleterious , the evolution of the number of loci @xmath5 in the genetic architecture is described by a markov - chain on the positive integers .
the probability of a transition from @xmath53 to @xmath54 equals @xmath55 , and that of a transition from @xmath6 to @xmath56 is @xmath57 .
we disallow transitions to @xmath58 , assuming that some regulation of the trait is required .
we obtained the stationary distribution of @xmath5 by setting the density of @xmath59 of individuals in stage @xmath60 to @xmath60 and calculating the density @xmath61 of individuals in the following stages as @xmath62 the equilibrium probability of being in state @xmath6 was calculated as @xmath63 and the expected value of @xmath5 was calculated as @xmath64 . with @xmath65 and @xmath66 , we found an equilibrium expected @xmath5 of @xmath67 .
when deletions , duplications and recruitments are all neutral , equation ( 4 ) can be replaced by : @xmath68
this equation illustrates the fact that the rates of deletions ( which include loss of function mutations ) and duplication depend on the number of loci in the architecture , whereas the rate of recruitments does not . with @xmath69 and @xmath70
, we found an equilibrium expected @xmath5 of @xmath71 .
we first evolved populations to equilibrium with a fixed number of controlling loci @xmath5 , and we then measured the effects of deletions , duplications or recruitments introduced randomly into the population .
we simulated the evolution of the genetic architecture with @xmath5 fixed in @xmath15 replicate populations , over @xmath74 generations for deletions and @xmath75 generations for duplications , reflecting the unequal waiting time before the two kinds of events .
we used @xmath75 generations for recruitment as well , although different durations did not affect our results .
for each genotype @xmath39 in each evolved population , we calculated the fitness @xmath76 of mutants with locus @xmath6 deleted or duplicated .
we calculated the corresponding selection coefficients as : @xmath77 where @xmath78 denotes mean fitness in the population .
we calculated @xmath72 as the mean across loci and genotypes of @xmath79 , weighted by the number of individuals with each genotype .
we calculated the probability of fixation of a duplication , deletion or recruitment as @xmath80 and obtained the mean @xmath81 using the same method as for @xmath72 .
rates of deletions and duplications fixing were calculated per locus ( fig .
2 ) as @xmath13 or @xmath12 times @xmath82 .
the total probability of a duplication or a deletion entering the population and fixing is , of course , also multiplied by @xmath5 .
however , recruitment rates remain constant as @xmath5 changes . therefore , we divided the rate of recruitments by @xmath5 in fig .
2 , for comparison to the per - locus duplication and deletion rates .
we used the _ r / qtl _ @xcite package to calculate lod scores for a set of @xmath83 observed markers and @xmath84 uniformly distributed pseudomarkers separated by @xmath85 cm , by haley - knott regression .
we calculated the lod significance threshold for a false discovery rate ( fdr ) of @xmath86 as the corresponding quantile in the distribution of the maximum lod after @xmath15 permutations ( a fdr of @xmath87 and a fixed lod threshold of @xmath88 produced qualitatively similar results ) .
the number of detected loci linked to the expression of a transcript was calculated as the number of non - consecutive genomic regions with a lod score above the threshold .
we downloaded _
s. cerevisiae _ coding sequences from the ensembl database ( ef3 release ) , and calculated cai values with the _ seqinr _ @xcite package , using codon weights from a set of @xmath89 ribosomal genes .
52 [ 1]#1 [ 1]`#1 ` ala - korpela , m. , a. j. kangas , and i. michaek , 2011 .
genome - wide association studies and systems biology : together at last .
trends genet . .
azevedo , r. b. r. , r. lohaus , s. srinivasan , k. k. dang , and c. l. burch , 2006 .
sexual reproduction selects for robustness and negative epistasis in artificial gene networks .
nature 440:8790 .
brem , r. b. and l. kruglyak , 2005 . the landscape of genetic complexity across 5,700 gene expression traits in yeast . proc . natl . acad .
u. s. a. 102:15721577 .
brem , r. b. , j. d. storey , j. whittle , and l. kruglyak , 2005 .
genetic interactions between polymorphisms that affect gene expression in yeast .
nature 436:7013 .
broman , k. w. , h. wu , s. sen , and g. a. churchill , 2003 .
r / qtl : qtl mapping in experimental crosses .
bioinformatics 19:889890 .
carlborg , o. , l. jacobsson , p. ahgren , p. siegel , and l. andersson , 2006 .
epistasis and the release of genetic variation during long - term selection .
38:418420 .
carter , a. j. r. , j. hermisson , and t. f. hansen , 2005 .
the role of epistatic gene interactions in the response to selection and the evolution of evolvability .
theor popul biol 68:17996 .
charif , d. and j. r. lobry , 2007 .
seqinr 1.0 - 2 : a contributed package to the r project for statistical computing devoted to biological sequences retrieval and analysis .
pp . 207232 , _ in
_ u. bastolla , m. porto , h. roman , and m. vendruscolo , eds . structural approaches to sequence evolution : molecules , networks , populations , biological and medical physics , biomedical engineering .
springer verlag , new york .
978 - 3 - 540 - 35305 - 8 .
conant , g. c. and a. wagner , 2004 .
duplicate genes and robustness to transient gene knock - downs in caenorhabditis elegans .
271:8996 .
ehrenreich , i. m. , j. bloom , n. torabi , x. wang , y. jia , and l. kruglyak , 2012 .
genetic architecture of highly complex chemical resistance traits across four yeast strains .
genet 8:e1002570 .
emilsson , v. , g. thorleifsson , b. zhang , a. s. leonardson , f. zink , j. zhu , s. carlson , a. helgason , g. b. walters , s. gunnarsdottir , m. mouy , v. steinthorsdottir , g. h. eiriksdottir , g. bjornsdottir , i. reynisdottir , d. gudbjartsson , a. helgadottir , a. jonasdottir , a. jonasdottir , u. styrkarsdottir , s. gretarsdottir , k. p. magnusson , h. stefansson , r. fossdal , k. kristjansson , h. g. gislason , t. stefansson , b. g. leifsson , u. thorsteinsdottir , j. r. lamb , j. r. gulcher , m. l. reitman , a. kong , e. e. schadt , and k. stefansson , 2008 .
genetics of gene expression and its effect on disease .
nature 452:4238 .
ewens , w. , 2004 .
mathematical population genetics : theoretical introduction .
springer verlag .
, a. and p. d. keightley , 2007 .
the distribution of fitness effects of new mutations .
8:610618 .
fierst , j. l. and t. f. hansen , 2010 .
genetic architecture and postzygotic reproductive isolation : evolution of bateson - dobzhansky - muller incompatibilities in a polygenic model .
evolution 64:67593 .
flint , j. and t. f. c. mackay , 2009 .
genetic architecture of quantitative traits in mice , flies , and humans .
. force , a. , m. lynch , f. b. pickett , a. amores , y. yan , and j. postlethwait , 1999 .
preservation of duplicate genes by complementary , degenerative mutations
. genetics 151:15311545 .
gu , z. , l. m. steinmetz , x. gu , c. scharfe , r. w. davis , and w .- h .
li , 2003 .
role of duplicate genes in genetic robustness against null mutations .
nature 421:6366 .
hansen , t. f. , 2006 .
the evolution of genetic architecture .
37:123157 .
innan , h. and f. kondrashov , 2010 . the evolution of gene duplications : classifying and distinguishing between models .
nat rev genet 11:97108 .
jones , a. , s. arnold , and r. brger , 2004 . evolution and stability of the g - matrix on a landscape with a moving optimum .
evolution 58:16391654 .
kimura , m. , 1985 .
the role of compensatory neutral mutations in molecular evolution .
64:719 .
kroymann , j. and t. mitchell - olds , 2005 .
epistasis and balanced polymorphism influencing complex trait variation .
nature 435:9598 .
lande , r. , 1976 .
the maintenance of genetic variability by mutation in a polygenic character with linked loci .
26:221235 .
lipinski , k. j. , j. c. farslow , k. a. fitzpatrick , m. lynch , v. katju , and u. bergthorsson , 2011 . high spontaneous rate of gene duplication in caenorhabditis elegans .
curr biol 21:30610 .
lynch , m. and a. force , 2000 .
the probability of duplicate gene preservation by subfunctionalization . genetics 154:459473 .
lynch , m. , w. sung , k. morris , n. coffey , c. r. landry , e. b. dopman , w. j. dickinson , k. okamoto , s. kulkarni , d. l. hartl , and t. w. k. , 2008 . a genome - wide view of the spectrum of spontaneous mutations in yeast . proc .
u. s. a. 105:92729277 .
mackay , t. f. c. , e. a. stone , and j. f. ayroles , 2009 .
the genetics of quantitative traits : challenges and prospects .
10:56577 .
manolio , t. a. , f. s. collins , n. j. cox , d. b. goldstein , l. a. hindorff , d. j. hunter , m. i. mccarthy , e. m. ramos , l. r. cardon , a. chakravarti , j. h. cho , a. e. guttmacher , a. kong , l. kruglyak , e. mardis , c. n. rotimi , m. slatkin , d. valle , a. s. whittemore , m. boehnke , a. g. clark , e. e. eichler , g. gibson , j. l. haines , t. f. c. mackay , s. a. mccarroll , and p. m. visscher , 2009 .
finding the missing heritability of complex diseases .
nature 461:74753 .
meer , m. v. , a. s. kondrashov , y. artzy - randrup , and f. a. kondrashov , 2010 .
compensatory evolution in mitochondrial trnas navigates valleys of low fitness .
nature 464:279283 .
ohno , s. , 1970 .
evolution by gene duplication .
london : george alien & unwin ltd .
berlin , heidelberg and new york : springer - verlag .
van ommen , g .- j .
b. , 2005 .
frequency of new copy number variation in humans .
37:333334 .
orr , h. a. , 1999 .
the evolutionary genetics of adaptation : a simulation study .
res . 74:20714 . ,
the genetic theory of adaptation : a brief history .
6:11927 .
poon , a. and s. otto , 2000 .
compensating for our load of mutations : freezing the meltdown of small populations .
evolution 54:14671479 .
proulx , s. r. , 2012 .
multiple routes to subfunctionalization and gene duplicate specialization .
genetics 190:73751 .
proulx , s. r. and p. c. phillips , 2006 .
allelic divergence precedes and promotes gene duplication .
evolution 60:881892 . , 2011 .
r : a language and environment for statistical computing .
r foundation for statistical computing , vienna , austria .
3 - 900051 - 07 - 0 .
rajon , e. and j. masel , 2011 .
evolution of molecular error rates and the consequences for evolvability .
usa 108:10821087 .
rieseberg , l. h. , m. a. archer , and r. k. wayne , 1999 .
transgressive segregation , adaptation and speciation .
heredity 83:363372 .
rockman , m. v. , 2012 .
the qtn program and the alleles that matter for evolution : all that s gold does not glitter .
evolution 66:117 .
rockman , m. v. and l. kruglyak , 2006 .
genetics of global gene expression .
7:86272 .
rockman , m. v. , s. s. skrovanek , and l. kruglyak , 2010 .
selection at linked sites shapes heritable phenotypic variation in c. elegans .
science 330:3726 .
rokyta , d. , m. r. badgett , i. j. molineux , and j. j. bull , 2002 . experimental genomic evolution : extensive compensation for loss of dna ligase activity in a virus .
biol . evol .
19:230 .
sharp , p. and w. li , 1987 .
the codon adaptation index - a measure of directional synonymous codon usage bias , and its potential applications .
nucleic acids res .
15:1281 .
ungerer , m. c. , s. s. halldorsdottir , j. l. modliszewski , t. f. c. mackay , and m. d. purugganan , 2002 .
quantitative trait loci for inflorescence development in _
arabidopsis thaliana_. genetics 160:11331151 .
visscher , p. m. , 2008 . sizing up human height variation .
40:489490 .
de visser , j. a. g. m. and s. f. elena , 2007 .
the evolution of sex : empirical insights into the roles of epistasis and drift .
nat rev genet 8:139149 .
wagner , a. , 2005 .
energy constraints on the evolution of gene expression .
biol . evol .
22:136574 . , 2007
energy costs constrain the evolution of gene expression .
308b:3224 .
wagner , g. , g. booth , and h. bagheri - cahichian , 1997 . a population genetic theory of canalization .
evolution 51:329347 .
watanabe , y. , a. takahashi , m. itoh , and t. takano - shimizu , 2009 .
molecular spectrum of spontaneous de novo mutations in male and female germline cells of drosophila melanogaster .
genetics 181:103543 .
yang , j. , b. benyamin , b. p. mcevoy , s. gordon , a. k. henders , d. r. nyholt , p. a. madden , a. c. heath , n. g. martin , g. w. montgomery , m. e. goddard , and p. m. visscher , 2010 .
common snps explain a large proportion of the heritability for human height .
nat genet 42:5659 . * supplementary material *
as described in the main text , the non - monotonic relationship between the strength of selection on a trait and the number of loci in its underlying genetic architecture depends on the unequal fitness consequences of deletions and duplications .
this behavior should therefore be absent when the trait value @xmath4 is the sum , rather than the average , of the contributions across loci . to explore this issue , we generalized our definition of the trait by introducing an additional parameter @xmath90 : @xmath91 as the parameter @xmath90 ranges from 0 to 1 the trait definition ranges from a sum to an average . as expected when @xmath92 , the number of loci in the equilibrium genetic architecture is greatly reduced under intermediate selection as compared to the results in the main text ( fig . [
fig : epsilon ] ) .
interestingly , the mean number of loci also shows a non - monotonic trend in this situation , with a small peak at @xmath93 .
this trend is likely driven by the fixation rate of recruitment mutations ( as seen in fig .
this explains why it is much less pronounced than those in figs . 1 and s3-a , which involve a higher fixation rate of duplications over deletions .
for any other value of @xmath94 , we found a non - monotonic relationship similar to the one reported in the main text .
thus , our qualitative results hold for all models provided the trait is not defined strictly as the sum of contributions across loci .
when the trait value equals the sum of the contributions of all locus ( @xmath92 ) , the effect of a gene deletion , knock - out or knock - down is independent of the number of copies of the gene .
conversely when the trait value is the mean of the contributions ( @xmath95 ) , or is some function between the mean and the sum ( @xmath96 ) , the effect of a deletion decreases with the number of loci in the genetic architecture . as shown by @xcite in @xmath97 , the number of detectable knock - down phenotypes decreases with the number of copies of genes in a gene family , suggesting that @xmath90 does indeed exceed @xmath8 in this species .
a similar stronger effect of the deletion of a singleton compared to that of a duplicate has also been observed in _ s. cerevisiae _ @xcite .
we analyzed the genetic architectures that evolved under our population - genetic model using a simulated qtl study of the exact same type and power as the yeast eqtl study @xcite .
specifically , @xmath98 evolved populations were taken from simulations with parameters corresponding to fig .
s2 for the model with epistasis . from each population
, we evolved two lines independently for @xmath99 generations in the absence of deletions , duplications and recruitment .
we then used the most abundant genotype from each line to create parental strains , mimicking the diverged by and rm parental strains in @xcite .
a few populations were polymorphic for the number of loci initially , sometimes resulting in two lines with different values of @xmath5 , which we discarded .
in each parent , we assigned the @xmath5 contributing loci randomly among @xmath83 simulated marker sites , and also assigned their associated @xmath7 values and the interactions @xmath27 between loci .
we constructed @xmath28 recombinant haploid offspring by mating these two parents according to the genetic map inferred from brem et al .
each offspring inherited each @xmath7 value , and the set of interactions towards other loci ( @xmath27 @xmath100 ) , from either one or the other parent .
the trait value in each offspring was calculated as in eq .
( 1 ) and then was perturbed by adding a small amount of noise ( normally distributed with mean @xmath8 and standard deviation @xmath101 ) , to simulate measurement noise .
we then analyzed these artificial genotype and phenotype data following the same protocol we used for the real yeast eqtls data ( i.e. using rqtl ) .
we repeated this entire process with @xmath98 different pairs of parents for each value of @xmath9 .
s18 shows the relationship between the selection pressure @xmath9 and the number of linked loci detected in this simulated qtl study , for different divergence times between the two lines and different values of @xmath101 . in fig .
s19 , we increased @xmath101 proportionally to @xmath102 , from @xmath103 to @xmath104
. we also calculated the probability that a locus known to influence the trait in the true architecture ( fig .
s3 ) is in fact detected in the qtl study .
this probability is plotted as a function of @xmath9 for different values of the noise @xmath101 ( fig .
s16 ) and of the time of divergence ( fig .
we simulated the evolution of the genetic architecture underlying multiple traits with a model slightly modified from the single - trait version . in this model ,
the phenotype consists of @xmath105 traits , each trait @xmath39 under a different selection pressure @xmath106 ( the values of @xmath106 are those used in independent simulations of the single - locus model ; see the x - axis of fig .
s15 ) . in the multiple traits version
, @xmath5 denotes the total number of loci forming the architecture of the @xmath105 traits .
@xmath5 can change when loci are duplicated at rate @xmath12 and deleted at rate @xmath13 .
each locus participates to a set of traits .
the direct effect of locus @xmath6 on trait @xmath107 is now denoted @xmath108 and the indirect effect of locus @xmath6 on the part of locus @xmath17 that contributes to @xmath107 is denoted @xmath109 . to allow for partial gains and losses of function ,
we define two new matrices @xmath110 and @xmath111 , which have the same dimensions as @xmath112 and @xmath113 . the functions corresponding to @xmath108 and @xmath109 are ` on ' when @xmath114 or @xmath115 , respectively , and are ` off ' otherwise . similarly to eq .
( 1 ) in our single - trait model , we calculate the trait value @xmath107 as : @xmath116 where @xmath20 is the sigmoidal function defined in eq ( 3 ) .
point mutations of locus @xmath6 alter all @xmath108 and @xmath109 by a normal deviate .
moreover , a mutation can change @xmath117 and @xmath118 to @xmath8 with probability @xmath119 and to @xmath60 with probability @xmath120 . over successive generations , the genetic architecture underlying each trait evolves through gene deletions and duplications , and through recruitments and losses of new functions . in this model ,
only the @xmath5 genes in the simulated architecture can be recruited
_ i.e. _ we do not assume a fixed number of genes that can be recruited at any time .
therefore , the phenotypic effects of recruitment evolve during our simulation , instead of being sampled from a given distribution .
if @xmath121 for any trait @xmath107 , the individual is considered non - viable and fitness @xmath122 equals @xmath8 . otherwise , fitness is the product of gaussian functions for each trait times the cost associated to the number of loci , as follows : @xmath123 we simulated the evolution of the genetic architecture through a wright fisher process , with population genetics parameters identical to the default values in table s2 , except @xmath124 @xcite ) .
the results of @xmath125 simulations are represented in fig .
s15 .
.estimates of rates of mutations @xmath11 , gene duplications @xmath12 and deletions @xmath13 .
all rates are per gene per generation .
@xmath11 is the rate of non - silent mutations @xcite ( @xmath126 the per - nucleotide mutation rate ) .
when the mutation rate was given per nucleotide , we multiplied it by the average gene length in eukaryotes @xcite ( @xmath127 bp ) . for _
d. melanogaster _
@xcite , the rate of detectable mutations was used , after correcting for the length of the @xmath88 loci in the study .
the scale of analysis can be the whole genome ( wg ) , or a specific set of loci , in which case the number of loci is denoted in the table . [
cols="<,^,^,^,^,^,>",options="header " , ] , depends on the strength of selection on the trait , in the model without epistasis .
a : traits under intermediate selection ( intermediate values of @xmath9 ) have more variable effects . under strong selection ,
the variance across loci is low because mutations changing the trait value are eliminated shortly and can not be compensated by other mutations . under weak selection , variance can increase through compensatory evolution when the architecture includes multiple loci , but this variance goes to @xmath8 when the number of loci reaches @xmath60 .
this occurs often enough ( see the distributions in fig .
1 ) to strongly reduce the mean variance of the phenotypic effects across loci .
parameters are set to their default values ( table s2 ) .
b : this difference across traits under various strengths of selection is also apparent in the distribution of @xmath7 .
the genetic architectures of traits under strong selection , and to a lesser extent of traits under weak selection are dominated by loci with small individual effects .
traits under intermediate selection rely on loci with more diverse contributions . ] ) evolve genetic architectures with the greatest number of controlling loci .
the rectangle areas are proportional to the number of wright - fisher simulations ( among @xmath15 per value of @xmath9 ) in which the number of loci on the y - axis evolved .
dots denote the ensemble mean of each distribution .
the neutral expectations for the equilibrium number of loci ( see methods ) are represented as grey lines , when recruitment events are neutral ( top line ) or not ( bottom line ; deletions and duplications are neutral in both cases ) .
the black arrow represents the number of loci for which the number of deletions fixing approximately equals that of duplications or recruitments for @xmath23 ( fig .
2 ) , where the mode of the distribution is expected .
b : standard deviations of @xmath7 ( direct effects ) and @xmath22 ( indirect ) are maximum under intermediate selection .
parameters are set to their default values ( table s2 ) . ]
at which the expected number of loci @xmath5 is maximum .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) . ]
increases with the mutation rate @xmath11 .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) .
a mutation rate of @xmath131 was used to sample recruitment events , so the overall probability of recruitment remains constant . ]
decreases as deletions become more frequent ( _ i.e. _ @xmath13 increases ) .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) . ]
increases as the probability of recruitment , @xmath26 , increases .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) . ] , has similar effects as decreasing @xmath11 .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) . ]
is not strongly affected by mutation biases in @xmath112 or @xmath113 .
a strong bias ( @xmath132 ) reduces the maximum variation across loci and therefore reduces @xmath5 when @xmath133 .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) .
one data point was omitted : @xmath134 at @xmath135 and @xmath136 . ]
decreases with the metabolic cost @xmath33 .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) . ]
whenever @xmath90 is higher than @xmath8 ( text s1 ) .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) . ]
is not affected by the form of the distribution of mutation effects .
all values represent the ensemble average of @xmath15 replicate simulations run for @xmath130 generations .
all other parameters are set to their default values ( table s2 ) . ]
loci and variable effects ( see description in the methods section ) . compared to fig .
s3-a , only the architectures of traits under strong selection have changed .
the mean number of loci under strong selection would be expected at the bottom grey line if deletions and duplications had neutral effects .
instead , this initial variation across loci prevents deletion or duplication , so the mean number of loci remains close to its initial value .
architectures of traits under intermediate and weak selection are not affected . ] | in the classic view introduced by r.a .
fisher , a quantitative trait is encoded by many loci with small , additive effects .
recent advances in qtl mapping have begun to elucidate the genetic architectures underlying vast numbers of phenotypes across diverse taxa , producing observations that sometimes contrast with fisher s blueprint . despite these considerable empirical efforts to map the genetic determinants of traits ,
it remains poorly understood how the genetic architecture of a trait should evolve , or how it depends on the selection pressures on the trait .
here we develop a simple , population - genetic model for the evolution of genetic architectures .
our model predicts that traits under moderate selection should be encoded by many loci with highly variable effects , whereas traits under either weak or strong selection should be encoded by relatively few loci .
we compare these theoretical predictions to qualitative trends in the genetics of human traits , and to systematic data on the genetics of gene expression levels in yeast .
our analysis provides an evolutionary explanation for broad empirical patterns in the genetic basis of traits , and it introduces a single framework that unifies the diversity of observed genetic architectures , ranging from mendelian to fisherian . * the evolution of genetic architectures underlying quantitative traits * + etienne rajon@xmath0 , joshua b. plotkin@xmath1 + @xmath2 department of biology , university of pennsylvania , philadelphia , pa 19104 , usa + @xmath3 e - mail : [email protected] + a quantitative trait is encoded by a set of genetic loci whose alleles contribute directly the trait value , interact epistatically to modulate each others contributions , and possibly contribute to other traits .
the resulting genetic architecture of a trait @xcite influences its variational properties @xcite and therefore affects a population s capacity to adapt to new environmental conditions @xcite . over longer timescales ,
genetic architectures of traits have important consequences for the evolution of recombination @xcite , of sex @xcite and even reproductive isolation and speciation @xcite .
although scientists have studied the genetic basis of phenotypic variation for more than a century , recent technologies , as well as the promise of agricultural and medical applications , have stimulated tremendous efforts to map quantitative trait loci ( qtl ) in diverse taxa @xcite .
these studies have revealed many traits that seem to rely on fisherian architectures , with contributions from many loci @xcite , whose additive effects are often so small that qtl studies lack power to detect them individually @xcite .
other traits , however , are encoded by a relatively small number of loci including the large number of human phenotypes with known mendelian inheritance .
the subtle statistical issues of designing and interpreting qtl studies in order to accurately infer the molecular determinants of a trait are already actively studied @xcite . nevertheless , distinct from these statistical issues of inferences from empirical data , we lack a theoretical framework for forming _ a priori _ expectations about the genetic architecture underlying a trait @xcite .
for instance , what types of traits should we expect to be monogenic , and what traits should be highly polygenic ? more generally , how does the genetic architecture underlying a trait evolve , and what features of a trait shape the evolution of its architecture ? to address these questions we developed a mathematical model for the evolution of genetic architectures , and we compared its predictions to a large body of empirical data on quantitative traits . |
the question of how dense , massive groupings like the trapezium are formed has yet to find a convincing answer .
ignoring for the moment the striking multiplicity of the system @xcite , a first - order glance at the trapezium shows the most massive stars in the cluster arranged in a central , compact configuration .
observations of a possible proto - trapezium , w3 irs 5 @xcite , seem to show a massive subcluster still in the embedded phase , which could be suggestive of either formation as a compact cluster or gas - driven migration during formation .
however , if a compact system of massive stars were to form _ in situ _ , it would be dynamically unstable @xcite ; if the stars in the trapezium formed in such a fashion , perhaps an order of magnitude more stars were originally in the system in order to leave behind the 4 ob stars today .
alternatively , the stars may have migrated there by some combination of dynamical mass segregation or gas dynamical effects during formation .
the problem with dynamical segregation alone ( as opposed to gas - driven segregation ) is the young age of the cluster and the segregation timescale , which taken together at face value would suggest that the cluster is not old enough to have segregated to the degree seen @xcite .
recent work @xcite has suggested a scenario in which merging smaller-_n _ subclusters can create a situation in which the final larger-_n _ cluster appears to be dynamically much older than the age of the stars would suggest .
the key concept is that dynamical mass segregation occurs on a shorter timescale in the low-_n _ subclusters , and that the segregation is maintained in the merger process , so that effects of the shorter dynamical timescale of the subclusters are applied to the structure of the final merger products .
this idea is supported by phase - space arguments by @xcite .
this sort of later dynamical mixing is perhaps supported by observations suggesting that young massive stars are frequently found in close proximity to other stars of varying ages @xcite .
the mcmillan et al .
results point the direction toward an attractive solution to the rapid concentration of massive stars as seen in the trapezium and w3 irs 5 . however , their simulations employed large numbers of stars , with @xmath0@xmath1 .
it remains untested how applicable these results are to less prolific but more representative star clusters with @xmath21000 , which may be more representative of a ` typical ' star forming region @xcite . in these smaller clusters with grainier potentials and crossing times that are smaller than their larger siblings ,
it is unclear if the factor of @xmath3 decrease in the dynamical time gained by halving the number of stars will come into play . in this paper
we investigate the merger of low-_n _ sub - clusters to determine if this process is likely to play a role in accelerating the concentration of massive stars in young local clusters .
there are several ways to measure the level of segregation in a cluster , including the evolution of the lagrangian radii , the half - mass radii of different mass components , or their cumulative number distributions .
a new method was recently developed by @xcite , utilizing the minimum spanning tree ( mst ) of the stars of interest relative to that of a random subset of the cluster stars .
an attractive feature of this method is that it does nt rely on defining the center of a cluster .
the method we use here is directly inspired by the mst method .
rather than use the mst , we use the two - dimensional convex hull of the stars positions projected onto the plane .
the convex hull encloses the minimal area that contains all straight lines joining any two points ; in two dimensions it is conceptually found by stretching an elastic band around the points of interest and letting it relax , and it is unique for a given set of points .
the area of the convex hull of @xmath4 stars , which we denote @xmath5 , divided by @xmath4 yields an effective surface density for the stars , and can be used like the mst to measure the compactness of a subset of stars in a cluster . for the given @xmath4 , and the progressively lighter gray regions
show the @xmath6 , @xmath7 , and @xmath8 regions .
the coloured lines show 10 realizations of a random setup for the plummer and ubr clusters , and for the onc case the data from @xcite .
the plummer spheres and ubr clusters all have 1000 stars and half - mass radii @xmath9 pc . [ convexhullexamples],width=317 ]
our scheme proceeds much as the mst method , in the following manner for the @xmath4 most massive stars in the cluster : first , compute @xmath5 for many subsets of @xmath4 stars drawn randomly from the cluster .
computing the convex hull is not intensive for @xmath10 points .
this work uses the built - in function implemented in _ mathematica_. ] ; @xmath11 subsets of 100 stars can be computed in @xmath12 seconds on an aging apple macbook with a 2 ghz intel core 2 duo and a loud cooling fan . after computing 1000 or more values of @xmath5
, the median and @xmath6 , @xmath7 , and @xmath8 values of the distribution can be determined .
these define the ` normal ' range of values for @xmath5 in the cluster for a _ random sampling _ of @xmath4 stars .
second , compute @xmath5 for the @xmath4 _ most massive _ stars . finally , repeat these steps for many values of @xmath4 up to the highest interesting value .
the compactness of the @xmath4 most massive stars can then be compared to the expected value for that @xmath4 , to see if they are more closely clustered than a random placing would be likely to yield . .
red dots show the positions of the eight most massive stars in the experiment.[mergeexample],width=317 ] to illustrate this scheme , we use it to compute the segregation levels for three test scenarios : an unsegregated plummer sphere ; a segregated cluster using the method of @xcite , using their value @xmath13 corresponding to a dynamically mature cluster ; and following @xcite , the orion nebula cluster using the data from @xcite . the plummer sphere and segregated clusters all have 1000 stars , and characteristic radii @xmath9 pc .
these examples are shown in figure [ convexhullexamples ] . in each panel ,
the black line is the median value of @xmath5 for 3000 random samples of @xmath4 stars , and the progressively lighter gray regions show the @xmath6 , @xmath7 , and @xmath8 values of this distribution .
these regions are to be compared to @xmath5 of the @xmath4 most massive stars , which we plot as coloured lines . looking at the plummer sphere in the top panel ,
note that most of the coloured lines , showing the values of 10 individual random realizations of the cluster setup , lie within the @xmath14 shaded region of typical surface densities for each @xmath4 .
this is then the indication of an unsegregated cluster .
note that the median values and shaded regions are taken from a single realization ; for these clusters and for the segregated cluster , the spatial distributions of each cluster are very similar and the median and shaded regions are almost indistinguishable between realizations . in the middle panel
we show a heavily segregated cluster , and we see that for all cases @xmath5 of the @xmath4 most massive stars lies above the @xmath15 value up to @xmath16 , and for seven out of the ten this extends out past @xmath17 .
this means that for most of the realizations of this setup , at least the 100 most massive stars are distributed in a fashion that is more compact than a random sampling of 100 stars would yield .
turning to the lower panel , we show the method as applied to the data from @xcite for the onc .
we see that for the lowest value of @xmath4 that we consider , @xmath18 , the surface density of the most massive stars is quite high . for @xmath19
, @xmath5 lies around or above the @xmath20 value of the cluster distribution , before falling back at or below the cluster median for higher values of @xmath4 .
this agrees quite well with the result @xcite found for the same data , which is unsurprising given that this method is fundamentally indebted to theirs .
note also the higher values of the surface density for the onc compared to the randomly generated clusters .
this illustrates the retention of physical meaning this method offers , showing both the absolute surface density and the relative amount of mass segregation for the most massive stars in a cluster .
we performed experiments to compare the amount of mass segregation obtained with a single plummer sphere to that resulting from the merger of two smaller plummer spheres .
the systems were integrated using nbody6 @xcite .
as these are merely illustrative experiments , we used a simple one - component mass function with masses in the range 0.1 10.0 m@xmath21and a power - law index @xmath22 .
we ran cases with two final cluster sizes : larger clusters where the single cluster had @xmath23 and scale length @xmath24 pc , and the merging clusters had @xmath25 and @xmath26 pc ; and smaller clusters where the single sphere had @xmath2 and @xmath27 pc , and the merging clusters had @xmath28 and @xmath29 pc .
these cluster sizes are a factor of a @xmath3024 times more compact then the clusters in @xcite(see figure 2 of * ? ? ?
* ) . if one considers the two merging spheres as a single cluster at a pre - merger time , however , they would be larger than that sample .
this is also ignoring the ( unmodeled here ) effects of gas dispersal and non - sphericity .
we chose the upper mass limit of 10 m@xmath31as we are interested primarily in low - to - moderate mass star formation .
additionally , our clusters are of relatively low mass ( @xmath32325 m@xmath21 ) , where one would arguably not expect to find a star of much greater mass than 10 m@xmath31@xcite .
the merging clusters were created by setting up one plummer sphere and setting its virial ratio , then duplicating the cluster .
the copies were initially at rest with respect to each other , placed so that they merged at an ` early ' time , @xmath33 yr , and a ` late ' time , @xmath34 yr . segregation was then measured at @xmath35 yr and @xmath36 yr for both the merger product and the single plummer sphere .
an example of this is shown in figure [ mergeexample ] . as there is some evidence that star formation may result in sub - virial stellar velocities ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , each setup was run from virialized and cold initial conditions , with virial parameters @xmath37 and 0.1 .
each set of parameters was run with 10 random realizations of the initial conditions . as a simple way to exclude escaping stars from the mass segregation calculation
, we restrict our analysis of the results to stars that remain within 1.5 pc of the center of mass .
we have anchored our experiments in physical time from the commencement of the integrations , rather than in terms of the initial crossing time , which only has observational significance to the single plummer spheres with their relatively static structure .
we therefore also determine for each set of parameters an _ apparent _ dynamical age , which is what a present - day observer might measure ( assuming reasonably accurate stellar ages can be determined ) .
we calculate this by determining the crossing time at the _ end _ of the experiment and dividing this by the age of the system in years , i.e. @xmath38 .
we define the crossing time as @xmath39 , with @xmath40 the half - mass radius and @xmath41 the root mean squared velocity .
we also show the age of the system in terms of its final relaxation time , @xmath42 , using the standard calculation for the relaxation time @xmath43 .
the mean value for each of these quantities for all sets of parameters is shown in table 1 .
the results of these experiments are shown in figures [ 1000synthesisvirial][300synthesiscold ] , where we plot the segregation up to the 60 most massive stars for each series of runs .
note that in these plots the median and its spread are calculated for a single realization , and all ten individual realizations are compared to this one .
while there are small differences in the median distribution between runs , for clarity we show only one , and this does not affect our conclusions . in table 1 , we summarize these plots and show the fraction of runs ( out of 10 ) @xmath44 that display segregation of the @xmath45 most massive stars , for @xmath46 . for these purposes
we define ` displaying segregation ' as having a value of @xmath5 above the @xmath15 value for the typical cluster value .
we also plot these fractions for the @xmath47 cluster in figure [ tableexample ] .
examining these figures and the table , it is clear that within a pair of runs with the same age , _
n _ , and @xmath48 the cluster resulting from a merger displays more segregation , with the difference between solitary plummer spheres and the merger products being more marked among the higher _
n _ clusters .
as we have chosen to fix the integration time to physical times rather than initial crossing times , this is unsurprising , since the lower-_n _ clusters are dynamically more mature .
the difference between the plummer spheres and the merger products is less pronounced with sub - virial initial conditions , and interestingly the @xmath47 and @xmath2 sub - virial plummer spheres have identical apparent dynamical ages at the end of the simulations .
it is also apparent that none of the initial conditions we examine here lead to a cluster displaying the full segregation of the clusters created using the technique of @xcite , which corresponds to the end - state of dynamical mass segregation .
the segregation in these simulations dominantly affects the most massive stars , a situation that looks more like the onc data from @xcite in figure [ convexhullexamples ] .
these results would then seem to confirm the findings of @xcite in the lower-_n _ regime that may be more applicable to local star formation . as the early evolution of young clusters with imposed primordial segregation across all masses appears to be inconsistent with young observed systems @xcite , this scenario is attractive since it seems to affect only the most massive stars in a cluster over these timescales and population levels .
a final point is that the single plummer sphere models with cold initial conditions exhibit more segregation of the most massive objects than their virialised counterparts ; while not at the same level as the merged clusters , this is another possible route to the formation of dense trapeziumesque systems .
examining the pairs of runs of the same physical age in terms of their apparent dynamical age , the general trend is that the merger product appears to be 14 dynamical times more advanced than the solitary plummer sphere .
the exceptions to this are the late - merging sub - virial case with @xmath47 and the late - merging virial case with @xmath2 , both of which appear to be dynamically younger than their solitary counterpart , though the difference is very slight . the sub - virial initial conditions for the @xmath47 plummer spheres actually appear to be dynamically older than their virial , merged counterparts , showing that both merging and sub - virial initial conditions can make a cluster appear older than its true age the fraction of runs showing segregation is not directly related to the apparent dynamical age , however ; the snapshot afforded by an observation may not reveal the history that led to that point
. in the sub - virial plummer sphere runs , for instance , the evolution of @xmath49 rises sharply from 0 to @xmath50 and plateaus for several hundred thousand years , as the cluster collapses , overshoots virial equilibrium , and expands . considering the apparent relaxation times at the end of the simulations
, one would not expect complete segregation as a result of two - body relaxation , as none of the systems are older than @xmath51 . since the segregation timescale for stars of mass @xmath52 decreases as @xmath53 , however @xcite
, one would expect to see some segregation in the massive stars based on the apparent relaxation times .
the fact that , for instance , the virial plummer sphere with @xmath47 and its late - merging counterpart have very similar ages relative to the relaxation time , yet very different mass segregation characteristics , highlights the limited applicability of the relaxation time in young systems of unknown history . in more complicated star formation scenarios ( i.e. non - spherical , or out of virial equilibrium ) , observational estimates of the crossing time or relaxation timescale should be treated with caution , as also emphasized by @xcite and @xcite .
a cautionary note should be sounded about the interpretation of these results and similar early - cluster stellar dynamical experiments .
there have been several attempts in the literature to apply _
n_-body techniques to very young clusters , when a proper treatment should necessarily include the gas reservoirs out of which the stars are born . to the extent that the remant gas is treated ,
it is usually as a spherical or spheroidal background potential on the cluster scale ( e.g. * ? ? ?
* ; * ? ? ?
* ) , or as an attempt to interpolate simulated hydrodynamic results on the level of individual stars @xcite . as it becomes more clear from observations and simulations that cluster formation may not be well - described as a spherical process ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , work from the _
n_-body end has pushed toward clustered and asymmetrical initial conditions as in @xcite .
it is not immediately clear that divorcing the stellar dynamics from the gas dynamics and treating effects like mass segregation as purely dynamical processes should be succesful ; the inclusion of gas dynamics as in the large - scale simulations as in , for example , @xcite or @xcite may prove to be obligatory . the background potential of any remaining gas can be included in an _
n_-body simulation with relative ease .
two immediately relevant dissipative effects that can not be modeled , but are present in hydrodynamic simulations , are the gas drag acting on the stars as they orbit in the cluster potential , and the mass loading that occurs in a competitive accretion scenario . both of these effects scale up with the mass of the star , and both should act to accelerate mass segregation . interestingly , @xcite reported minimal segregation in his recent simulation that included these effects as well as merging sub - clusters , indicating that ( as is usually the case ) matters are more complicated than simple prescriptions might allow for .
this suggests that while effects like deceleration due to bondi - hoyle accretion could perhaps be added to an _
n_-body code , the utility of such approximations may be limited .
a comparison between pure _ n_-body merging experiments can be made to the recent hydrodynamic/_n_-body simulation presented in @xcite . in this simulation ,
the cluster at @xmath54 yr is the end result of the merger of five sub - clusters consisting of @xmath55 stars .
these mergers took place in the presence of the cluster gas .
bate notes that mass segregation is not present in the end cluster as measured by the cumulative radial distribution , or the median mass , upper quartile mass , and maximum mass except for in the inner 1000 au .
applying the segregation detection method used in this paper to the positions of the stars in bate s simulation , we find evidence of segregation for the most massive stars , shown in figure [ batecluster ] . while there is clearly not general segregation as in the ubr clusters , the @xmath56 most massive stars display a level of segregation similar in character to the onc data , and would not stand out as out - of - place in amongst the simulated data in figures [ 1000synthesisvirial ] and [ 1000synthesiscold ] .
while the issue of the equivalence between gas - free and embedded stellar dynamics should be explored further , in this case the level of segregation seen appears to be similar in character .
the mass segregation signal resulting from the merger of sub - clumps of an age and stellar content suitable to nearby young clusters appears to affect predominantly the most massive stars in the cluster , and does not present itself as a general trend that might appear in a cumulative distribution analysis or similar techniques .
full - scale hydrodynamic simulations @xcite that feature merging sub - clusters display a similar signal , despite the mergers taking place in the presence of gas .
while the more general type of segregation may have been looked for as the ` goal ' of segregation - producing phenomena in the past , the type that affects only the most massive stars may be more reflective of nature ( as seen in the onc ) , and thus sub - cluster merging and perhaps sub - virial initial conditions seem to be a promising scenario .
@xcite published the results of an independent investigation that overlaps significantly with this work .
namely , they studied the dynamical effects of sub - virial , clumpy collapse on the mass segregation in a cluster .
their initial cluster is a more naturalistic setup versus the stylized two - plummer - sphere clusters we consider here , with 1000 stars placed via a fractal distribution .
the rapid segregation seen in the @xcite experiments is mostly due to two - body relaxation in the dense core that develops as the sub - clusters merge and the cluster collapses at early times .
the extent to which pre - merger segregation in the sub - clusters contributes to the final state ( mcmillan et al .
2007 ; fellhauer et al . 2009
; this work , see figure [ mergeexample ] ) versus two - body relaxation in the dense core resulting from clustered , sub - virial collapse @xcite is likely dominated by the details of the initial setup .
finally , we caution against an overly - broad definition of trapezium - like or trapeziumesque .
figure [ trapezium ] shows the locations and clustering dendrogram of the four ob systems in the trapezium and their entourage @xcite .
most work , including this one , focuses on the four main branches of the dendrogram , i.e. attempting to create a clustering of four or more massive stars .
while this is clearly the first step toward a trapezium , there are at least two more levels of clustering , in the mini - trapezium of @xmath57 and the wider binary of the @xmath58 hierarchy , and then the closer visual and spectroscopic binaries of all four main components .
turning to the younger w3 irs 5 , with the further complication of surrounding natal gas , there is evidence for multiplicity at sub-1000 au scales @xcite .
bearing in mind the limited number of examples , multiplicity of the massive sources in trapezium - like systems seems ubiquitous , and could be fundamental rather than a detail to be worried about later . if dynamical effects like those considered here and in @xcite are to explain the clustering of massive stars , these processes must either leave intact the presumably primordial multiplicity of the massive stars , including higher order subsystems like @xmath57 , or produce the multiplicity as part of the segregation ; neither possibility has yet been investigated , and consistency with this aspect of the trapezium may help discriminate between dynamical segregation versus birth location as the cause of massive star clustering .
our thanks to matthew bate for sharing the stellar positions from his simulation , and the referee for a prompt review . | the nearest site of massive star formation in orion is dominated by the trapezium subsystem , with its four ob stars and numerous companions .
the question of how these stars came to be in such close proximity has implications for our understanding of massive star formation and early cluster evolution .
a promising route toward rapid mass segregation was proposed by @xcite , who showed that the merger product of faster - evolving sub clusters can inherit their apparent dynamical age from their progenitors . in this paper
we briefly consider this process at a size and time scale more suited for local and perhaps more typical star formation , with stellar numbers from the hundreds to thousands .
we find that for reasonable ages and cluster sizes , the merger of subclusters can indeed lead to compact configurations of the most massive stars , a signal seen both in nature and in large - scale hydrodynamic simulations of star formation from collapsing molecular clouds , and that sub - virial initial conditions can make an un - merged cluster display a similar type of mass segregation .
additionally , we discuss a variation of the minimum spanning tree mass - segregation technique introduced by @xcite .
[ firstpage ] methods:_n_-body simulations stars : formation stellar dynamics |
supermassive black holes ( smbhs ) appear to be present in the nucleus of most , and perhaps all , nearby galaxies ( see , e.g. , reviews by @xcite and @xcite ) .
the correlations between the masses of the smbhs and various global properties of the host galaxies suggest that evolution of smbhs is closely related to the evolution of galaxies .
in particular , in hierarchical structure formation models , galaxies are built up by mergers between lower mass progenitors .
each merger event is expected to deliver the nuclear smbhs ( e.g. @xcite ) , along with a significant amount of gas @xcite , to the central regions of the new post merger galaxy .
there is some evidence for nuclear supermassive black hole binaries ( smbhbs ) , which would be expected to be produced in galaxy mergers .
direct x ray imaging of an active nucleus
@xcite has revealed a smbh binary at a separation of @xmath10kpc , and @xcite recently identified a candidate smbhb , at @xmath11 times smaller separation , from its optical spectrum .
a radio galaxy is also known to have a double core with a projected separation of @xmath12 pc @xcite , and several other observations of radio galaxies , such as the wiggled shape of jets indicating precession ( e.g. * ? ? ?
* ) , the x shaped morphologies of radio lobes ( e.g. * ? ? ?
* ; * ? ? ?
* ) , the interruption and recurrence of activity in double
double radio galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ) , and the elliptical motion of the unresolved core of 3c66b @xcite below . ] have all been interpreted as indirect evidence for smbh binaries down to sub pc scales .
two interesting conclusions may be inferred from the above observations .
first , while there is evidence for a handful of nuclear smbhbs , these objects appear to be rare .
this suggests that if binaries do form frequently , then they coalesce ( or at least their orbital separation decays to undetectably small values ) in a small fraction of the hubble time .
second , smbhbs can apparently produce bright emission , with a luminosity comparable to active galactic nuclei ( agn ) , before they coalesce . in general , the circumbinary gas , delivered to the nucleus in galactic mergers ,
can both play a catalyst role in driving rapid smbhb coalescence @xcite , and could also accrete onto one or both smbhs , accounting for bright emission during the orbital decay .
the dense nuclear gas around the bh binary is expected to cool rapidly , and settle into a rotationally supported , circumbinary disk ( e.g. * ? ? ?
* ; * ? ? ?
the dynamical evolution of a smbhb embedded in such a thin disk has been studied in various idealized configurations ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the generic conclusion of these studies is that initially , the orbital decay is relatively slow , and is dominated by viscous angular momentum exchange with the gas disk , whereas at small separations , the decay is much more rapid , and is eventually dominated by gravitational wave ( gw ) emission . whether the decaying smbhbs produce bright electromagnetic ( em ) emission is comparatively much less
well understood .
if the disk is thin , the torques from the binary create a central cavity , nearly devoid of gas , within a region about twice the orbital separation ( for a nearly equal mass binary , e.g. * ? ? ?
* ) , or a narrower gap around the orbit of the lower mass bh in the case of unequal masses @xmath13 ( e.g. * ? ? ?
* ) . in the latter case , the lower mass hole `` ushers '' the gas inward as its orbit decays , producing a prompt and luminous signal during coalescence . in the former case ,
if the central cavity were indeed truly empty , no gas would reach the smbhbs , and bright emission could not be produced .
however , numerical simulations suggest residual gas inflow into the cavity @xcite , which may plausibly accrete onto the bhs , producing non negligible em emission .
yr after the coalescence , caused by the gas outside the cavity falling in , after a delay set by the disk viscous time @xcite .
such an afterglow is interesting , for example for a follow up to smbh merger events detected by the _ laser interferometric space antenna ( lisa ) _ , but not relevant to the idea proposed in the present paper . ] finally , smbhbs recoil at the time of their coalescence due to the emission of gravitational waves @xcite
. the gas disk will respond promptly ( on the local orbital timescale ) to such a kick , which may produce shocks , and transient em emission , after coalescence @xcite .
the kick , however , can begin building up during the late inspiral phase @xcite , possibly resulting in some emission even before the final coalescence . during the late stages of coalescence , emission may also be produced by viscous heating of the disk by the gws themselves @xcite .
the luminosity , spectrum , and time evolution of any em emission produced by coalescing smbhbs , especially during the last , gw driven stages , remains uncertain .
_ however , any emission produced during the inspiral stage is likely to be variable .
_ for example , recent numerical simulations of an equal mass binary on parsec scales @xcite , and of both equal and unequal mass binaries on sub parsec scales @xcite find that the circumbinary gas disk is perturbed into eccentric orbits by the rotating quadrupole potential of the binary , and that the rate of residual accretion across the edge of the cavity is modulated , tracking the orbital period .
the luminosity is likely to be directly tied to the mass accretion rate , and therefore may vary periodically .
however , even if the gas accretion rate were steady , one would expect periodic flux variations , due to the orbital motion of the binary ( kocsis & loeb , in preparation ) . in this paper
, we address the question : _ given their expected rate of orbital decay , could the population of coalescing smbhbs be identified statistically in an observational survey for periodically variable sources ? _ given that the interpretation of individual smbhb candidates have so far remained ambiguous , with alternative explanations possible for each source , the potential for such a statistical identification should be explored . to answer this question , we first utilize steady state thin disk models to study the orbital decay of a smbhb , embedded in a circumbinary disk .
the decay is described by the residence time @xmath14 the sources spend at each orbital radius @xmath15 , or at the corresponding orbital timescale @xmath16 . in the limiting case of a purely gw driven evolution , which becomes valid at small orbital separations ( typically at @xmath17 several@xmath18 schwarzschild radii , but with large variations ; see below ) and remains valid until the final minutes of the merger ( the so called `` plunge '' stage ) , the residence time is given by @xmath19 . at larger separations , the viscous interaction between the binary and the disk drives the binary evolution .
the residence time in this regime becomes dependent on assumptions about the properties of the disk and the nature of the binary
disk interaction , which we will explore in this paper . in general , @xmath20 , with the generic value of @xmath21 well below @xmath22 significantly flatter than the @xmath23 relation in the gw driven stage .
we then hypothesize that ( i ) non negligible emission ( at a fair fraction of the eddington luminosity ) is maintained throughout the orbital decay , and ( ii ) the luminosity varies periodically on the orbital time
the first assumption allows us to identify coalescing smbhbs with luminous quasars .
the second assumption implies that as the orbit of a binary decays , its variability timescale decreases . among sources at redshift @xmath24 with
similar inferred bh masses , the observed incidence rate @xmath25 of periodic variability on the time
scale @xmath26 , is then proportional to the residence time @xmath27 . at short periods
, the @xmath25 could therefore show a characteristic power
law dependence on @xmath28 indicative of a gw driven evolution , whereas at longer periods ( and , as we will discuss , for lower bh masses ) the dependence will be flatter , due to viscosity driven evolution .
we quantify the requirements that such periodically variable sources be identifiable , based on their incidence rate , in an optical or x ray survey .
luminosity variations at a fraction @xmath29 of the eddington luminosity would correspond to a periodically varying flux component with amplitude @xmath30 for bhbs at @xmath31 , or to @xmath32 $ ] magnitudes in the optical .
we find that these periodic sources are either too faint or too rare to have been found in existing variability surveys .
however , if the overall luminosity is indeed a non negligible fraction of the binary s eddington luminosity , then a long duration future survey , sensitive to periods of weeks to tens of weeks , could look for periodically variable sources , and identify a population of sources obeying well defined scaling laws .
the discovery of a population of such periodically variable sources could have several implications . at long periods and low bh masses ,
the scaling index @xmath21 between the residence time and the period @xmath33 will probe the physics of the circumbinary accretion disk and viscous orbital decay . at shorter periods and higher masses ( roughly , at @xmath34 few weeks for @xmath35 ) , the identification of a @xmath36 power law would confirm that the orbital decay is driven by gws .
this would amount to an indirect , statistical detection of gw driven smbhbs , independent of any direct detection of gws by _
lisa_. this would also confirm that circumbinary gas is present at small orbital radii and is being perturbed by the bhs and would thus serve as a proof of concept for finding _ lisa _ electromagnetic counterparts .
the rest of this paper is organized as follows . in
[ sec : diskmodels ] , we discuss the evolution of binaries with different masses and mass - ratios , embedded in a circumbinary gas disk .
we describe simplified models for the disk and for the binary disk interaction , and emphasize that the binaries probe the distinct physical regimes in the disk , before gws take over their evolution . in [ sec : observations ] , we discuss the possibility of searching for a population of coalescing smbhbs among a catalog of luminous quasars , either based on their variability , or on shifts of their spectral lines .
we discuss modest constraints available from existing surveys , and comment on specific recently detected individual smbhb candidates .
we then quantify the requirements for a detection in a future survey . in
[ sec : summary ] , we briefly summarize our results and offer our conclusions .
when necessary in this paper , we adopt the background cosmological parameters @xmath37 , @xmath38 , and @xmath39 @xcite .
in this section , we describe the evolution of the orbital separation of a smbh binary . the basic picture
we adopt is that the binary is embedded in a thin circumbinary disk , with the plane of the disk aligned with the binary s orbit @xcite .
initially , the orbital decay is dominated by viscous angular momentum exchange with the gas disk . however , the time
scale for viscous decay decreases relatively slowly as the orbital separation @xmath15 decreases ( @xmath40 ; see below ) whereas the time scale to decay due to gravitational radiation decreases steeply ( @xmath41 ) .
therefore , generically , there exists a critical orbital radius @xmath42 , below which the decay is dominated by gravitational radiation . to describe the evolution quantitatively ,
we make several simplifying assumptions . the circumbinary gas is assumed to form a standard geometrically thin , optically thick , radiatively efficient , steady state accretion disk @xcite . we assume zero eccentricity for both the binary and for
the disk ( justified by @xcite , however see @xcite ) , and we assume co planarity between the disk and the binary @xcite .
all of these assumptions may fail in the late stages of the merger ( even before gw driven decay begins ) . however , under these assumptions , the disk structure and the orbital decay have simple limiting power law solutions , with the power
law indices depending on the choice for the underlying physics . these solutions are useful to describe the possible evolution of the binary , and to illustrate the point that the decay rate is generically a different much flatter
function of @xmath16 than the @xmath43 behavior in the gw driven case .
we emphasize that our aim here is not to provide accurate , self consistent solutions for the co evolution of the smbh binary and circumbinary disk .
rather , we derive only gross scaling laws in various regimes our main point is that these regimes and associated uncertainties , which are large , can in principle be probed observationally .
we adopt the following notation throughout this paper .
we refer the reader to @xcite and @xcite for general introductions to accretion disks . * _ physical constants : _
@xmath44 is the gravitational constant ; @xmath45 is the speed of light ; @xmath46 is the boltzmann constant ; @xmath47 is the stefan boltzmann constant ; @xmath48 is the thompson cross section ; @xmath49 is the mean mass per electron in units of hydrogen atom mass , @xmath50 , which satisfies @xmath51 for a fully ionized gas of both hydrogen and helium ; @xmath52 is the mass fraction of hydrogen ; @xmath53 is the mean molecular weight ; @xmath54 is the electron scattering opacity ; and @xmath55 ( t/{\rm k})^{-7/2}$ ] is the rosseland mean absorption opacity in the free free regime @xcite . *
_ bh parameters : _ @xmath56 and @xmath57 are the individual bh masses ; @xmath58 is the total bh mass ; @xmath59 is the mass ratio ; @xmath60 is the normalized symmetric mass ratio ; @xmath61 is the reduced mass ; @xmath15 is the binary separation ; @xmath62 is the location of the lower mass secondary , measured from the center of mass of the binary ; @xmath63 is the schwarzschild radius corresponding to the total mass ; @xmath64 is the eddington luminosity for a bh of mass @xmath0 ; @xmath65 is the eddington accretion rate with a radiative efficiency @xmath66 ; and @xmath67 is the characteristic time scale associated with eddington accretion . * _ disk parameters : _
@xmath68 is the vertical scale height ( the effective geometrical semi
thickness of the disk ) ; @xmath69 is the volumic gas density ; @xmath70 is the surface density ; @xmath71 is the gas pressure ; @xmath72 is the radiation pressure ; @xmath73 is the total pressure ; @xmath74 ; @xmath75 is the ( midplane ) gas temperature ; @xmath76 is the effective temperature defined such that the locally emitted flux through an infinitesimal disk surface element is @xmath77 ; @xmath78 is the keplerian orbital angular velocity ; @xmath79 is the outer radius of the gap in the punctured circumbinary disk , measured from the center of mass of the binary ; @xmath80 is the anomalous dynamical viscosity ; @xmath81 is the anomalous kinematic viscosity ; @xmath21 is the standard viscosity parameter of thin accretion disks ; @xmath82 is a constant , either 0 or 1 , determining whether viscosity scales with the total or just the gas pressure , so that @xmath83 ; @xmath84 is the opacity of the disk material ; @xmath85 is the vertical optical depth ; and @xmath86 is a constant defined such that @xmath87 .
quantities with a @xmath88 subscript ( e.g. @xmath89 ) denote parameters in a steady state disk around a single unperturbed accreting bh , computed at the radius @xmath79 .
similarly , quantities with a 0 subscript are evaluated at the position of the secondary @xmath90 . with the above definitions , we proceed to define the dimensionless quantities @xmath91 , @xmath92 , @xmath93 , @xmath94 , @xmath95 , @xmath96 , @xmath97 , @xmath98 , @xmath99 , and @xmath100 .
note that radii are measured from the center of mass of the binary .
we adopt the set of fiducial values @xmath101 , @xmath102 , @xmath103 , @xmath104 , is appropriate for a one zone model where all the energy is dissipated near the midplane and the opacity is constant vertically @xcite . for reference , we note that @xcite and @xcite adopt different values of @xmath105 and @xmath106 , respectively . ] and denote values relative to these fiducial choices with a tilde , e.g. @xmath107 .
our fiducial binary+disk model is therefore chosen to be @xmath108 .
since all of our expressions can be written as products of power laws in the physical parameters , the resulting expressions become tractable in these units .
we next collect the basic expressions from the literature for accretion disk models under different physical conditions .
we quote the equations for a range of different steady thin disks , valid for a single accreting bh @xcite .
we distinguish several cases : _
( i ) _ whether the radiation or gas pressure provides the dominant vertical support , _
( ii ) _ whether the opacity is dominated by electron scattering , @xmath109 , or free free absorption , @xmath110 , and _ ( iii ) _ whether the viscosity @xmath80 is proportional to the total pressure or the gas pressure ( also known as @xmath21 and @xmath111 disk models , respectively ) . based on these choices
, the accretion disk can be divided radially into three distinct regions
@xcite : 1 . _
inner region : _ radiation pressure and electron - scattering opacity dominate , @xmath112 , @xmath113 , valid inside @xmath114 where @xmath115 is defined in equations ( [ e : gas / radb1 ] ) and ( [ e : gas / radb0 ] ) below . 2 .
_ middle region : _ gas pressure and electron - scattering opacity dominate , @xmath116 , @xmath117 , valid between @xmath118 , where @xmath119 is defined in equation ( [ e : es / ff ] ) below .
_ outer region : _ gas pressure and free - free opacity dominate , @xmath116 , @xmath120 , valid outside of @xmath121 . in region
( 1 ) , it makes a difference whether the viscosity is proportional to the total pressure or just the gas pressure , labeled below by @xmath122 or 1 , ( i.e. @xmath21 or @xmath111 disk ) respectively . in all cases , we assume that the disk is optically thick , i.e. @xmath123 .
we obtain @xmath124 and @xmath125 following @xcite or @xcite , @xmath126 where @xmath122 or 1 , and the radial dependence is implicit in @xmath127 and @xmath111 . here ,
@xmath128 which satisfies @xmath129 the asymptotic limits of equations ( [ e : sigma_definition ] ) and ( [ e : h_definition ] ) can be obtained in regions ( 13 ) , using equation ( [ e : beta_definition ] ) .
the results are _ inner region : _ @xmath130 _ middle region : _ @xmath131 _ outer region : _
@xmath132 the boundaries between the inner / middle and middle / outer regions can be found from equations ( [ e : sigma_definition])-([e : beta_definition ] ) , by requiring @xmath133 and @xmath134 , respectively .
note that @xmath135 depends on radius implicitly through the density and the temperature . using the ( mid - plane ) temperature given by @xcite , @xmath136
we find that the transitions are located at the radii @xmath137 note that the middle and outer regions differ only in their opacity laws , and the equations in these two regions are equivalent ( this can be seen by setting @xmath138 ) . since @xmath139 , @xmath68 , @xmath69 , and @xmath75 scale with a low power of @xmath140 , the radial dependence ends up being similar in the middle and outer regions .
the distinction between these equations is nevertheless useful , since we can assume that @xmath141 and @xmath142 are constants in the middle and outer regions , respectively .
we emphasize that equations ( [ e : sigma_inb1])-([e : h_out ] ) represent only a very non - exhaustive subset of solutions even for radiatively efficient steady thin accretion disks .
in particular , at large radii , there are several effects that can invalidate the disk model described by these equations .
first , these solutions assume that the self
gravity of the disk is negligible .
this assumption becomes invalid at radii where the toomre @xmath143parameter equals unity , @xmath144 beyond these radii , the disk is commonly believed to be unstable to fragmentation .
second , at large radii , the disk can also become optically thin ( see * ? ? ?
* where solutions can be obtained by fixing the toomre parameter in the outermost region at @xmath145 ) . at these binary separations , the disks may not actually be geometrically thin @xcite , and slim or thick solutions might instead be relevant .
third , beyond the radii where the disk temperature falls below @xmath146k , the gas becomes neutral .
the corresponding change in opacity will modify the disk structure , and the disk may become susceptible to ionization instabilities ( although see @xcite ) . finally , at large radii ( where the orbital velocity @xmath147 km / s ) , the gravitational potential of the galaxy can no longer be ignored .
these regimes , however , turn out to correspond to separations larger than we are interested in the present paper , for bh masses above @xmath148 ( as will be shown in figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] below ) .
we have verified our solutions numerically by substituting them back into the fundamental conservation equations of thin accretion disks @xcite . moreover , equations ( [ e : sigma_inb1])-([e : h_middle ] ) agree with those quoted in @xcite . in their eq .
20 should be @xmath149 , and @xmath150 in their eq .
21 should be @xmath151 and @xmath152 in the inner and middle regions , respectively . ]
equations ( [ e : sigma_inb0])-([e : h_in ] ) are also consistent with @xcite , for the @xmath153 , @xmath154 , @xmath122 model .
it is also reassuring that equations ( [ e : sigma_out])-([e : h_out ] ) are consistent with those in ( * ? ? ?
8.1 , p. 244 ) . , which should instead read as @xmath155 , so that @xmath156 is satisfied for all @xmath157 .
] also note that , owing to the weak dependence on @xmath158 , our numerical factors are very similar to those in @xcite , even though @xmath110 is defined to be two orders of magnitude larger there than the value we adopted here ( to be consistent with most other textbooks ) . here
we collect and summarize the most important formulae describing the interaction between a binary and the accretion disk in order to identify the mechanism that drives the orbital decay of the binary during the final stages of the merger , as a function of binary separation ( the choices being gw driven inspiral and tidal viscous torques ) .
this will allow us to explicitly compute the residence time @xmath14 that an individual binary spends at each orbital separation @xmath15 , or at the corresponding orbital timescale @xmath16 .
the formulae collected in this section will also allow us to quantify the binary separation at which the viscous evolution of the disk is decoupled from the increasingly rapid , gw driven orbital decay of the binary .
we provide results for @xmath21 and @xmath111disks , and give analytic results as a function of binary and disk parameters . in general
, the evolution of a smbh binary in a thin disk is analogous to planetary migration ( see , e.g. @xcite ) . in the limit of a very low mass companion ( @xmath159 ) ,
the interaction between the planet and the disk is linear .
in addition to co - rotation resonances , the density waves excited in the gas at discrete lindblad resonances with the binary exert a large net torque on the binary , leading to rapid , so called type
i migration , which occurs on a time
scale much shorter than the local viscous time
scale ( e.g. , @xcite ) .
if the binary is massive enough for the tidal torque to dominate over the viscous torque in the disk , the interaction becomes non linear , and a gap is opened in the disk , extending to the outer radius @xmath160 .
the condition for a gap to open is that the mass ratio exceeds the critical value @xmath161 ( e.g. * ? ? ?
* note that @xmath162 is evaluated at the position of the secondary @xmath90 ) . for binaries that are not in the gw driven regime , and for which the disk mass exceeds the mass of the secondary ( see below ) , this typically translates into the very modest requirement @xmath163 .
this is satisfied for all smbh binaries that may produce the electromagnetic signatures we discuss below .
the exceptions are the so
called extreme mass ratio binary inspirals ( emri s ) with @xmath164 ( i.e. a stellar mass object coalescing with a smbh ) . in this paper , we focus on smbh binaries , and therefore in the rest of this paper , we neglect type i migration . if the secondary s mass satisfies the above gap opening threshold , but is still small compared to the local disk mass , then it acts as an angular momentum bridge for the disk , and the secondary s orbital evolution is simply determined by the viscous diffusion time , @xmath165 where we have used @xmath166 which follows directly from angular momentum conservation in steady disks @xcite .
the orbital decay of the binary in this limit is analogous to _ disk dominated type - ii planetary migration_. in practice , the assumption that the local disk mass exceeds the secondary s mass often fails . in this case ,
analogous to _ `` planet dominated '' type - ii migration _ , the angular momentum of the binary can still be absorbed by the gaseous disk outside the gap , and the viscosity of the gas can drive the binary toward merger . however , migration is slower , and the time scale in this regime , @xmath167 , is longer than @xmath168 .
an estimate of the slowing factor is @xmath169 , where @xmath170 is a measure of the lack of local disk mass dominance ( * ? ? ?
* but note that our @xmath171 is denoted by
@xmath172 in their original definition ) , which is less than unity in this case , and @xmath173 is a constant defined as @xmath174 thus , the separation of the binary in this case is driven inward on the timescale @xmath175 note that the viscous time
scale @xmath168 in disk dominated limit ( eq . [ e : armitage_natarayan ] ) should be evaluated at the position of the secondary @xmath62 , while the quantities entering the time scale @xmath167 for the secondary dominated type - ii migration of more massive binaries ( @xmath176 , eq .
[ e : t_ii ] ) , should be evaluated at the outer edge of the cavity , @xmath160 @xcite . in order to avoid a discontinuous jump in the migration time
scale at the @xmath177 transition , below we will omit this distinction , and evaluate both time
scales at @xmath178 . using the steady thin disk model outlined above
, we can calculate the rate at which the binary is driven inward by the gas .
we will also estimate the rate at which the inner edge of the punctured gaseous disk follows the binary due to its viscosity . from the preceding discussion
, we see that both the viscous time scale and the orbital decay rate depend on whether the binary is located in the inner / middle / outer region of the disk ; and also on whether the local disk mass is larger / smaller than the mass of the smaller smbh . for completeness ,
we here obtain and quote the residence time as a function of orbital radius and orbital time , in each of these @xmath179 regimes .
we then construct the self consistent evolution of individual binaries , with different masses and mass - ratios , across the relevant regimes .
we first consider the timescale @xmath168 , and assume that the secondary perturbs the disk at the radius @xmath180 .
this is the relevant regime initially , at large binary separations , when the disk mass enclosed within the secondary s orbit is large . in this regime , we find , @xmath181 the above can be expressed as a function of the orbital time of the binary , @xmath182 which results in @xmath183 the measure of disk dominance can be calculated by substituting the viscous time
scale into equation ( [ e : q_b ] ) , @xmath184 in order to decide whether the evolution indeed follows the disk dominated decay ( on the viscous timescale @xmath168 ) or the secondary dominated decay ( on the longer time scale @xmath167 ) , one should examine whether @xmath185 or @xmath176 is satisfied , respectively . from equations ( [ e : q_b-1])-([e : q_b-4 ] ) , we find that the transition occurs at @xmath186 note that with the exception of very unequal masses @xmath187 , the transition takes place well in the outer region of the disk , with @xmath188 . at smaller radii ,
the binary is driven viscously on the timescale @xmath167 ( rather than @xmath168 ) .
the `` secondary dominated ''
type - ii decay timescales relevant at these radii can be obtained by substituting equations ( [ e : t_nur-1])-([e : t_nur-4 ] ) into equation ( [ e : t_ii ] ) @xmath189 or , in terms of @xmath16 using equation ( [ e : t_orb ] ) , @xmath190 finally , at a still smaller radius , the orbital decay will be dominated by gravitational wave emission .
the gw driven decay timescale in the leading order ( newtonian ) approximation , is @xmath191 this approximation is adequate for our purposes , since post
newtonian corrections do not become appreciable until the final @xmath192 day of the merger ( see , e.g. , figure 5 in @xcite ) .
note that @xmath193 defined above differs from the total time to merger , ( defined as the binary separation decreasing to zero ) , which is often used in the literature , and which occurs at @xmath194 .
what is the radius at which @xmath193 becomes smaller than the time
scale for type - ii orbital decay ?
let us express this transition in terms of the radius @xmath195 that satisfies @xmath196 , where @xmath197 is a fixed constant of order unity : @xmath198 the corresponding critical radius is around @xmath199 for system parameters near the assumed fiducial values . the critical radius , however , is significantly closer in for very massive , and very unequal mass binaries ( i.e. for @xmath200 and @xmath201
; see fig .
[ fig : tres_r_q0.01 ] below ) .
interestingly , the critical radius is quite insensitive to the bh mass and accretion rate ( i.e. to @xmath202 and @xmath203 ) .
note that the viscous timescale , @xmath168 , describing gas accretion , is faster than @xmath167 , which indicates that at the time when gw starts driving the evolution , the viscous inward diffusion of gas can initially still follow the binary .
however , the comparison of equations ( [ e : t_nur-1])-([e : t_nur-4 ] ) and equation ( [ e : t_gw ] ) shows that as the binary orbit shrinks further , the viscous time - scale always decreases less rapidly than the gw inspiral timescale , so that eventually the evolution of the gaseous disk will decouple from that of the binary .
let us find the critical radius , @xmath204 , where gw inspiral outpaces viscous gas accretion .
we find that in most cases , this critical radius is not relevant for the orbital decay of the bhs themselves , because the transition to secondary driven orbital decay always takes place before gws start dominating the decay .
, equal mass binaries , and only if @xmath205 is assumed in this case , the gw inspiral takes over in a radiation pressure dominated disk , in the disk dominated regime , i.e. before the transition to the secondary dominated regime .
] however , this critical radius is relevant for the behavior of the disk : it provides an estimate for the time when the punctured disk decouples from the gw driven binary , and effectively stops evolving ( and also for the size of the inner gap at this time and onward ) . by requiring @xmath206 , where @xmath207 is a constant coefficient of order unity ,
we obtain : @xmath208 the appropriate choices for @xmath209 and @xmath207 are poorly known , but @xmath210 may be reasonably taken to be @xmath211 when the binary is first driven by gw emission , rather than by tidal interaction with the gas .
the simplest choice for @xmath207 , adopted in many previous studies , is also @xmath212 ( e.g. * ? ? ?
* ; * ? ? ?
however , the gas inflow rate across the edge of the central gap will be increased due to the steep density and pressure gradient @xcite , which will delay the decoupling .
this motivated @xcite to adopt @xmath213 ( the value describing the limiting case of an infinitely sharp edge ) .
adopting @xmath214 in equations ( [ e : r_ii / gw1])-([e : r_ii / gw4 ] ) then yields the radius where the binary evolution changes from being viscosity driven to gw driven , and @xmath215 in equations ( [ e : r_nu / gw1])-([e : r_nu / gw4 ] ) gives the separation at which the disk totally decouples from the binary and the radius of the gap `` freezes '' .
these expressions generalize the results of @xcite , who restricted their analysis to the @xmath205 case , and focused on the behavior of the disk at decoupling , rather than the orbital evolution of the binary .
in particular , @xcite evaluate disk conditions at the single radius at the edge of the gap , at the time of decoupling , and do not discuss the transition from the disk to the secondary dominated decay , or other details of the binary s orbital decay .
the binary separation at decoupling is of order @xmath216 for both the gas pressure dominated models and the radiation pressure dominated case with @xmath205 . in these cases , the transition between viscosity and gw driven decay and the disk decoupling take place in relatively quick succession , since @xmath217 depends weakly on @xmath207 .
the delay between these two events is much longer for the radiation pressure dominated regime when @xmath122 , since in this case the viscosity , which is proportional to the total , rather than just the gas pressure , is much larger , and the gas can follow the binary nearly all the way to merger ( at least for large @xmath203 ) . in this case
, the result is also extremely sensitive to the accretion rate and the binary mass ratio .
generically , for a fixed total binary mass , the decoupling occurs at the largest separations for nearly equal masses .
interestingly , the decay rate of a given individual binary can decelerate and accelerate during its evolution , according to the variations in the local disk environment at each instantaneous binary separation .
the evolutionary tracks of binaries with four different choices for the total mass ( @xmath218 , and @xmath219 ) and two different mass ratios ( @xmath220 and @xmath201 ) are shown in figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] . in both figures , we assume that the viscosity is proportional to the total pressure ( @xmath122 ) . the motivation for this choice is to illustrate the effect of the additional radiation pressure related viscosity on the orbital decay ( which is not present in the @xmath205 case ) .
we note that a phenomenological @xmath122 disk is known to suffer from a formal thermal instability ( e.g. * ? ? ?
* ) ; recent magnetohydrodynamical simulations , however , found such disks thermally stable ( while accounting for the correlation between viscosity and radiation pressure * ? ? ?
these figures show the residence time as a function of the orbital time .
they demonstrate that the evolution of the binary in most cases proceeds through the following distinct stages . _
( i ) disk dominated viscous evolution . _ initially , at large separations ( shown in blue curves ) , the binary is strongly coupled to the circumbinary disk and evolves on the viscous time
scale @xmath221 ( analogous to `` disk - dominated '' planetary migration ) .
the radius of the gap follows the binary . during this stage
, @xmath222 is proportional to @xmath223@xmath224 ( the range corresponding to the choice @xmath122 _ vs. _ @xmath205 ) for radiation pressure , or @xmath223@xmath225 ( the range corresponding to the choice of dominant opacity being electron scattering or free free absorption ) for gas pressure dominated disks ( see eqs .
[ e : t_nur-1][e : t_nur-4 ] ) .
these decay rates translate into @xmath226@xmath227 , or @xmath228@xmath229 , in the two cases respectively ( see eqs . [
e : t_nuorb-1][e : t_nuorb-4 ] ) .
note , however , that for nearly equal mass binaries , the separations have to be quite large to correspond to this disk dominated regime falling into the outer regions of the disk , which are unstable to fragmentation ( the orbital radii where the disks are marginally toomre
stable are marked with large dots ) .
therefore , depending on the behavior of the gas disk beyond this radius , this early stage of disk dominated viscous evolution may exist only for unequal
mass binaries .
as shown in figure [ fig : tres_q0.01 ] , disk dominated viscous evolution may be realized in a stable disk for binaries with @xmath230 and @xmath231 ; in these cases , the binaries are in the free free opacity and gas pressure dominated regions of the disk , so the relevant scaling is @xmath232 . _
( ii ) secondary dominated viscous evolution . _ as the binary separation shrinks below @xmath233 ( @xmath234 ) for mass ratios @xmath235 ( @xmath231 ) , the binary mass starts to dominate over the local disk mass , and the binary evolves more slowly , according to `` secondary dominated '' decay ( analogous to `` planet - dominated '' type - ii migration ) . during this stage , the gw emission is still negligible , and the decay time
scale can be obtained from equations ( [ e : t_iir-1])-([e : t_iir-4 ] ) , and @xmath236@xmath237 for radiation pressure ( with @xmath238 ) , or @xmath239@xmath240 for gas pressure dominated ( with electron scattering vs. free free opacity ) disks , implying that @xmath241@xmath242 , and @xmath243@xmath244 , in the two cases respectively ( see eqs .
[ e : t_iiorb-1][e : t_iiorb-4 ] ) . as can be seen from figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] , on orbital time
scales between weeks to years , each of these scalings is relevant for some choice of binary masses .
however , the transition to gw domination always takes place either in the `` inner '' or `` middle '' disk region . _
( iii ) gw dominated evolution .
_ still later , within the radius @xmath245 for systems with parameters close to the fiducial values , the binary s orbital evolution starts to be driven primarily by gws , but the outer edge of the gap can still diffuse inward and follow the binary . during this stage , the decay time scale is @xmath246 . _
( iv ) gas disk decoupled .
_ finally , within @xmath247 the binary is entirely driven by gws and the binary falls in much more quickly than the outer edge of the gap is able to move inward . the above ordering of events
is valid for a broad range of binary and disk parameters .
note that the ultimate fate of the gas inside the binary s orbit is left unspecified in our considerations ( see , e.g. * ? ? ?
* for a possible outcome ) .
in addition to the above sequence of events describing the evolution of individual binaries , several interesting conclusions can be drawn from figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] . 1 .
_ coalescing binaries have a non - negligible abundance . _
first , binaries with masses in the range @xmath248 may be both bright and common enough to be detectable in a survey , provided they have bright emission . indeed ,
figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show that these binaries spend a non negligible fraction ( @xmath249 ) of their total fiducial lifetime of @xmath2 years at orbital time
scales between 1 day @xmath3 1 year ( the total lifetime will be justified below ) .
it is feasible , in principle , to look for variability on these time scales , and the residence times shown on the figures suggest that these variables may not be uncommon among bright agn .
we will discuss this possibility further in
[ sec : periodic ] below . 2 .
_ disk and gw driven evolution may both be observationally relevant .
_ figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] also show that the transition from gas to gw driven evolution can occur within this `` observational window '' .
for example , at the fixed orbital time scale of @xmath250 weeks , equal mass binaries above @xmath251 are gw driven , and below this mass , they are gas driven .
secondary dominated evolution can not be ignored .
_ essentially all binaries at the orbital times relevant for actual surveys ( again , between 1 day @xmath3 1 year ) that are gas driven are in the regime of `` secondary dominated '' type - ii orbital decay ( referred to as stage _
( ii ) _ above ) .
likewise , the transition from `` gas driven '' to `` gw driven '' evolution always occurs from the `` secondary dominated '' type - ii decay regime . in previous works whose primary focus was on the behavior of gas at ( and
after ) the time of decoupling ( e.g. @xcite ) , this intermediary step , which is important for the orbital decay of the binary , is not discussed .
4 . _ observed binaries could probe all three disk regions . _
interestingly , among the @xmath248 binaries with 1 day @xmath3 1 year , it appears that all three of the disk regions ( inner / middle / outer ) enumerated in
[ subsec : thindisks ] can be observationally relevant ( i.e. , gas driven binaries can be found in each of these three disk regions ) .
viscous evolution is non
negligible even in the lisa regime . _
the comparison of figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] shows that unequal
mass binaries evolve more rapidly when they are gas driven .
consequently , they make the transition to the gw driven stage quite late in their evolution . in particular , binaries enter _ lisa _ s detection range at the approximate observed gw frequency of @xmath252 mhz .
this corresponds to an observed orbital time ( on earth ) of @xmath253 week .
we find that at this orbital time , viscous evolution is not necessarily negligible .
figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show that equal
mass binaries with @xmath254 , and @xmath201 binaries with @xmath255 are just making the transition to the gw - driven regime as they enter the _ lisa _ band .
_ total decay time in a stable disk is consistent with quasar lifetime .
_ as figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show , the residence time at the radius at which @xmath256 is , in all cases , close to ( although somewhat shorter ) than the fiducial quasar lifetime of @xmath2 years . it is plausible that smbhs become luminous , and act as quasars , only once they are embedded in stable circumbinary accretion disk . the fact that it takes @xmath257 years for the binary to evolve from the outer edge of a stable disk to coalescence is therefore consistent with the idea proposed in this paper , that there is a one to
one correspondence between coalescing smbhs and quasars ( although , as mentioned above , there are caveats that can invalidate the steady disk models at the relevant large radii ) . the possible implication of conclusion no . 5 above for _ lisa _ merits some further elaboration .
as discussed , e.g. , in @xcite , individual binaries can contribute to the _ lisa _ data stream in several ways .
sources can be divided into two types , based on whether they evolve significantly on a time
scale of @xmath258 years , the duration of the _ lisa _ experiment .
binaries caught at an orbital separation with short enough residence times for the frequency evolution to be measurable are sometimes referred to as `` gravitational sirens '' or `` gravitational inspirals '' .
figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show that during the last several years of the coalescence , the orbital evolution is always strongly gw dominated , even for the lowest mass bhs , and therefore the gw waveform of these rapidly evolving sources ( including those whose actual coalescence is detected by _
lisa _ ) will not be affected by the gas disk .
binaries that have a much longer residence time at some fixed frequency in _
s band represent `` stationary '' sources whose frequency remains roughly constant during the _ lisa _ mission lifetime .
these sources could , in principle , be individually detectable by _
lisa_. however , in practice , they are likely to accumulate sufficient signal to noise for detection only in the last few hundred years of their coalescence ( see , e.g. , figure 2 in @xcite for the detectability of @xmath259 binaries as a function of their look back time from the merger ) . figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show that viscous processes can significantly speed up the evolution of binaries only at somewhat larger look back times ( note that the look
back time is 4 times shorter than the residence time in the pure gw driven case ) .
the cumulative signal from a collection of faint stationary sources can , however , still add up to an unresolved background that is detectable , depending on the the cosmic evolution of the bh merger rate and the instrumental noise of _ lisa_. the presence of the gas disks could reduce any such background that is present ( compared to a prediction that assumes pure gw driven evolution at _ lisa_frequencies ) . in figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] , we have showed the evolution of the binary as a function of its orbital period .
this will be particularly useful for assessing the detectability of such binaries in a survey for periodically variable sources ( [ sec : periodic ] below ) . in figures [ fig : tres_r_q1 ] and [ fig : tres_r_q0.01 ] , we show , instead , the evolution of the same set of binaries , but as a function of their orbital separation .
the @xmath260axis on these figures is shown in units of @xmath261 , with the corresponding orbital velocities shown by the labels on the top axis .
this figure directly reveals that relatively more massive binaries ( @xmath262 ) spend a significant time at orbital velocities of several thousand @xmath263 .
such orbital speeds may be detectable in the spectra of individual sources , providing an alternative to the detection based on periodic flux variations ( see [ sec : alternatives ] below ) .
finally , the conclusions enumerated above also highlight the large uncertainty in the residence times predicted in figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] , caused by our idealized treatment of `` secondary dominated '' type - ii orbital decay .
one immediate additional source of uncertainty is the choice of @xmath122 vs. @xmath205 . before entering the gw driven regime , most of the equal mass binaries ( fig .
[ fig : tres_q1 ] ) are in the gas pressure dominated region of the disk , but unequal mass binaries ( fig .
[ fig : tres_q0.01 ] ) are in the radiation pressure dominated region . therefore ,
whether the viscosity is proportional to the total pressure or just the gas pressure makes little difference to the near equal mass binaries .
however , it makes a significant difference for unequal
mass binaries with @xmath264 . to show this explicitly , in figure [ fig : tres_b_q0.01 ] , the upper vs. lower curves contrast the evolution in the @xmath205 vs @xmath122 case , respectively .
as expected , once the binary approaches the radiation pressure dominated regime , the evolution is significantly slower in the @xmath205 case .
the difference is most pronounced for the most massive ( @xmath265 ) binary . for this system
, the transition to gw domination also occurs at a larger orbital time ( @xmath266 weeks for @xmath205 , vs. @xmath267 weeks for @xmath122 ) . for simplicity , above we calculated the timescales @xmath168 and @xmath167 in steady thin disk models .
however , as noted above , this highly idealized model makes several crucial assumptions .
in particular , ( * ? ? ?
* hereafter ipp ) considered the tidal
viscous interaction of an unequal mass binary ( @xmath159 ) with a _ time - dependent _ accretion disk .
they assumed that the accretion disk is initially described by the steady state solution for a single bh , and then considered the modifications due to tidal torques from a secondary bh .
the torques are turned on suddenly at some moment @xmath268 , when the secondary , whose mass is @xmath269 , is at an orbital radius @xmath270 that encloses a disk mass @xmath271 .
the torques are assumed to be concentrated in a narrow ring near the secondary s orbit , which results in a pile up of material near the outer edge of the disk cavity .
they found ( see their eq .
58 ) that this results in a decay time scale of @xmath272 where @xmath273 is the initial steady state accretion rate , @xmath274 is the initial viscous time ( at @xmath275 ) , @xmath276 is the disk dominance parameter at @xmath275 ( see eq .
[ e : q_b ] ) , @xmath277 is the time - dependent position of the secondary and @xmath278 is its initial position , and @xmath279 ^ 2 $ ] , with the dimensionless time @xmath280 , implying that @xmath281 where @xmath282 , and @xmath283 , @xmath82 , and @xmath173 are defined such that @xmath284 , @xmath285 , @xmath286[(2c+a)/(2c+1)]^{-(2c+a)/(a+1)}$ ] , @xmath287 $ ] , and @xmath168 is the unperturbed viscous timescale given by equation ( [ e : armitage_natarayan ] ) and calculated explicitly below .
refers to the standard gas pressure dominated accretion disk , which we generalize to radiation dominated disks below . ] if the opacity is dominated by electron scattering , then @xmath288 , @xmath205 , @xmath289 , @xmath290 , and @xmath291 , while for the free - free process @xmath292 ,
@xmath293 , @xmath294 , @xmath295 , and @xmath296 . from equations ( [ e :
t_ipp ] ) and ( [ e : tau_ipp ] ) we find @xmath297 here , the radial evolution given by equation ( [ e : t_ivanov ] ) is ( at least initially ) not a simple power law .
most importantly , as noted by @xcite , the pile up of the disk material causes the binary decay to slow
down _ even more _ than estimated for a steady disk based on the `` disk dominance '' parameter ( eq . [ e : t_ii ] above ) . in figures [ fig : tres_ipp_q1 ] and [ fig : tres_ipp_q0.01 ] , we illustrate the impact of allowing the disk to evolve . in the above approach of @xcite
, we have to specify when the interaction between the secondary and the disk is turned on . in figure
[ fig : tres_ipp_q1 ] , we assume that the interaction begins as soon as the disk dominance parameter reaches @xmath177 . in figure
[ fig : tres_ipp_q0.01 ] , we delay the onset of the interaction to @xmath298 . in both figures , the new dotted ( magenta )
curves denote the binary s residence time in the time
dependent disk . note that in the latter case , in figure [ fig : tres_ipp_q0.01 ] , the residence time undergoes a discrete jump when the disk binary interaction is turned on : the binary stalls , and does not move initially , until the mass of material that has piled up is of the order of the secondary s mass .
as the figures show , these residence times are indeed significantly longer than in the steady disks . at relatively late times after the interaction is assumed to turn on , the residence times asymptote to power law forms .
most significantly , for each of the binaries shown in figures [ fig : tres_ipp_q1 ] and [ fig : tres_ipp_q0.01 ] , the transition to the gw driven regime occurs significantly earlier due to the evolution of the disk ; just before this transition to the gw driven regime , the residence times are longer by @xmath299 two orders of magnitude compared to a steady disk .
in the rest of this paper , we will discuss identifying coalescing smbhbs with quasars , and interpreting the residence time @xmath300 as the duty
cycle for exhibiting periodic variability on the observed time
scale @xmath301 .
our broad justification for these hypotheses is the generic idea , advanced in numerous other works , that quasars are activated in major galaxy mergers ( e.g. @xcite and references therein ) . since smbhs
are believed to be common in galactic nuclei ( at least at low redshifts @xmath302 ; see @xcite and references therein ) there could then arguably be a one to
one correspondence between the quasar phenomenon and smbhb coalescences . the total quasar lifetime , defined as the cumulative duration ( possibly over multiple episodes ) for an individual source to produce bright emission near the eddington limit , is generally believed to be @xmath303 yr , based on several lines of observational evidence @xcite . as discussed above , this value is consistent with the time
scale it takes for a binary smbh to evolve to coalescence , starting from the outer edge of a gravitationally stable thin @xmath21disk .
therefore , we hypothesize that the luminous quasar phase coincides with this last stage in the merger of the two smbhs . of course , it is possible that the quasar phase occurs either _ long before _ or _
after _ coalescence in either case , there would be no bright emission to observe during the last stages , as hypothesized here .
we next assume that during the coalescence , the binary produces a steady luminosity @xmath304 ( which evolves only on long time scales @xmath305 ) , with roughly periodic fluctuations of amplitude @xmath306 and period @xmath307 about this steady mean luminosity . as argued in the introduction , periodic variations could be reasonably expected if the luminosity is tied to the mass accretion rate , with the latter modulated on the orbital period . even in the absence of such modulations ,
the emission could vary owing to the orbital motion and emission geometry of the binary ( kocsis & loeb , in preparation ) . in the absence of a quantitative model for the electromagnetic emission
, we will assume the amplitude @xmath306 is unknown , and below we will ask whether a particular assumed @xmath306 may be detectable . since the residence time @xmath300 decreases continuously as the binary separations shrinks , variability with decreasing periods @xmath16 would be exhibited by a diminishing fraction @xmath308 of bright quasars .
_ our main point is that an observational survey can attempt to identify such periodically variable sources . _
the total number of such periodic sources will be @xmath309 , where @xmath310 represents the merger rate between bhs within the survey volume ( or more precisely , the activation rate of smbh coalescence events ) .
in general , the merger rate depends on redshift and on both bh masses , or @xmath311 , and should include only those sources with a luminosity above the survey detection threshold . to account for the latter condition , the light curve of each smbhb , @xmath312 needs to be known ( here @xmath313 could , for example , refer to the look back time before merger ) .
the merger rate @xmath314 can be modeled using the dark matter halo merger rate with a recipe of associating bhs with halos , and the light
curve @xmath315 can then be constrained by matching the observed quasar luminosity function ( e.g. * ? ? ?
however , a large range of such bh population models can fit the observational data ( e.g. * ? ? ?
* ; * ? ? ?
* ) . to proceed , we instead make the simple assumption that each bh binary produces a constant mean luminosity of @xmath316 for a total duration @xmath317 during its lifetime , where @xmath318 is a constant of order unity .
reasonable fiducial values appropriate to the bright quasar phase are @xmath319 @xcite and , as mentioned above , @xmath320 yr @xcite .
note , in particular , that the quasar lifetime @xmath317 is known to be much shorter than the hubble time , and @xmath310 , which is likely determined by the galaxy merger rate , and proceeds on a cosmological time scale , can reasonably be assumed to be constant during @xmath317 . under the above assumptions , the fraction @xmath25 of objects with luminosity @xmath321 that display periodic variability on the time - scale @xmath322
is simply given by the ratio @xmath323 .
this ratio can be read off directly from figures [ fig : tres_q1]-[fig : tres_ipp_q0.01 ] .
note that this conclusion still holds if the quasar emission is intermittent ; we require only that the quasar is `` on '' for the duration @xmath300 when the binary orbital timescale is @xmath16 .
most importantly , under these assumptions , the predicted number @xmath324 is a fixed fraction of the total number @xmath325 of quasars , and is independent of the merger rate , as long as the latter is constant during @xmath317 .
we can then associate @xmath325 with the _ observed _ number of bright agn .
in particular , in the gw dominated regime , we have the simple prediction @xmath326^{8/3 } m_{7}^{-5/3 } q_s^{-1}.\ ] ] note that in this equation , @xmath327 is the variability time - scale as observed on earth ( assumed to equal the redshifted orbital time ) ; the quasar lifetime @xmath328 is evaluated in the quasar s rest frame .
before we proceed , we emphasize that there are many complications over the above , simplified picture .
first , luminous quasar activity requires a near eddington mass accretion rate , with the gas reaching within several schwarzschild radii of one or both bhs .
it is unclear whether abundant gas will indeed be present this close to the bhs , especially since during the late stages of the merger , the gas is evacuated from the inner disk by the binary s torques , and the exterior gas disk is eventually unable to follow the rapidly decaying bh binary .
furthermore , in the final , gw dominated regime ,
the @xmath329 scaling strictly holds only if any residual circumbinary gas has negligible impact on the orbital decay .
this requirement could , in fact , contradict the assumption that the binary is producing bright emission during this stage .
second , in order for the emission to be periodically variable , the gas has to respond rapidly to the gravitational perturbations from the binary .
scale for this response is of order the _ local _ orbital time ; variability on the orbital time
scale of the _ binary _ itself therefore again requires gas close to the binary s orbital radius . if the central cavity were indeed truly empty , no gas would reach the smbhbs , and bright emission could not be produced . on the other hand ,
an empty cavity is certainly an idealization , and detailed models for the joint disk + binary evolution are required to assess the plausibility of our assumptions .
conversely , the observations envisioned here will constrain such models ( which , again , is the main point of the present paper ) . in support of our assumptions , we note , however , that gas could be present near the bhs in the case of unequal masses ( so that the torques are reduced ) , or if the disk remains thick , making it difficult for the binary to open and maintain a nearly empty central cavity .
numerical simulations indeed suggest residual gas inflow into the cavity @xcite , which may plausibly accrete onto the bhs ( with both bhs possibly forming their smaller individual accretion disks ; @xcite ) , producing non negligible em emission .
simulations have also shown , in the context of proto planetary disks , that when the circumbinary disk is sufficiently thick , the mass flow rate across the gap is increased @xcite . such residual inflow onto a smbh binary
has been invoked to explain the @xmath19212yr periodic emission from the quasar oj287 @xcite .
more recently , the large velocity offsets seen in the spectrum of the quasar sdss j092712.65 + 294344.0 @xcite have been interpreted with a similar model , including gas inflow onto a luminous smbh binary @xcite ; a similar interpretation was invoked for the binary quasar candidate recently identified by @xcite .
there are additional caveats that will hamper the identification of any periodic sources , even if they exist and produce bright enough luminosity to be detectable .
the eddington ratio of bright agn is already known to have a significant scatter ( @xmath330 dex ; @xcite ) .
the light
curve of the merging binary is also likely to evolve , rather than having a simple `` tophat '' shape .
it is possible , in particular ( e.g. * ? ?
* ) that merging smbhs spend a significantly longer time ( @xmath331 yr ) at lower luminosities , ( @xmath332 ) .
this will complicate the interpretation of any observed variability ( i.e. , converting the observed ratio @xmath333 at @xmath322 to @xmath334 will require knowing the probability distribution of eddington ratios ) .
this , however , can be alleviated by considering only the _ relative _ abundance of periodically variable objects at different values of @xmath322 , instead of the absolute number of sources that show periodic variability . in this case
, the only assumption required is that @xmath318 does not evolve significantly during the observed range of @xmath322 this should be reasonable over a factor of a @xmath192 few range in orbital radius or in @xmath16 .
furthermore , even if there is a range of different bh masses , among sources with a similar luminosity , producing variability with the same period , figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show that more massive bhbs will move much more quickly through a fixed @xmath335 .
given that there are most likely fewer of the more massive bhbs to begin with , the set of all sources with the same @xmath322 will be heavily dominated by the lowest mass bhbs , caught at their relevant orbital radius .
this still leaves the caveat , however , that the source is significantly sub eddington during the late stages of coalescence . in this case
, the periodic sources will be harder to detect both because they are fainter , and also because they will also be rarer ( among the long lived and therefore more numerous , near
eddington quasars with a similar luminosity ) .
another caveat is that at fixed @xmath16 and @xmath0 , the distribution of @xmath1 is unknown , and can depend on @xmath0 .
however , bright agn activity is thought to be activated only in relatively major mergers .
a smaller satellite galaxy , falling onto a larger central galaxy that is more than @xmath267 times more massive , may not experience the torques needed to bring its gaseous nucleus , with the low mass bh , close to the center of the larger galaxy , for the bh - bh merger to take place @xcite ; the dynamical friction time for small galaxies themselves can also be too long ( e.g. * ? ? ?
* ) , and/or the small satellites can be tidally stripped before reaching the central regions of the larger galaxy ( e.g. * ? ? ?
these arguments , coupled with the well established correlations between the mass of a smbh and its host galaxy ( e.g. @xcite , see also the introduction ) , suggest that the @xmath1distribution among binaries associated with quasars may not extend to values significantly below @xmath336 .
finally , for simplicity , in our estimates we have assumed circular orbits , both for the binary and the disk gas .
it has been shown that the binary
disk interaction could drive both the smbhs and the gas to have significant eccentricities @xcite .
such eccentricities should leave characteristic asymmetric signatures in the modulated mass accretion rate ( see figure 8 in @xcite ) . the resulting light
curves may exhibit corresponding features , which could be resolved , given sufficient time sampling . in practice , allowing for eccentricities will most likely further complicate the interpretation of any observed period distribution , especially if the time sampling is too coarse to explicitly reveal any asymmetric features . despite
the caveats listed in the previous section , it is plausible that the periodic sources envisioned here exist , and we propose that they can be looked for , in a suitably designed survey .
most importantly , figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show that the expected variability timescale can be in a suitable range for a statistical detection , with a duty cycle of @xmath337 yr over the range from @xmath5day to @xmath192yr .
this suggests that such periodic sources may not be too rare
. what will be the practical limitations for discovering the population of periodic sources ? clearly , there has to be a sufficient number of sources , observed over a range of variability time
scales for a representative statistical sampling , and the brightness variations of these sources must be detectable .
in addition , the individual light curves have to be sampled well enough to confirm their periodic nature : this will be necessary to distinguish the coalescing smbh binaries from other types of variable objects .
besides discovering the periodic sources , the idea proposed here is to measure the dependence of @xmath338 on @xmath322 possibly to use the @xmath339 scaling to demonstrate that the periodic variability comes from perturbations by the orbital motion during the gw inspiral . for this , the survey also needs to cover at least a factor of several range in @xmath322 .
the above issues will place requirements on ( i ) the sensitivity and ( ii ) solid angle , as well as on the ( iii ) total duration and ( iv ) sampling rate for a survey .
we can use the simple disk models and the idealized picture discussed above , to roughly delineate these requirements .
for simplicity of discussion , let us assume that all sources are at @xmath31 .
in reality , quasars ( and therefore major bh mergers ) have a broad distribution with a peak around this redshift ; clearly this will have to be taken into account in designing an actual survey .
for simplicity , let us also fix the mass ratio @xmath220 .
in reality , there should be a distribution of values , perhaps in the range @xmath340 , for the mergers that activate bright quasar activity @xcite .
this would not significantly affect our conclusions , unless @xmath1 frequently extends well below 0.1 .
imagine a survey with a sensitivity that corresponds to detecting the periodic variability of bhbs with a mass @xmath341 at @xmath31 , covering a solid angle @xmath342 .
( a real survey , of course , will have a completeness for variability detection that is not a step function ) .
let us assume that the variable flux corresponds to a fraction @xmath343 of the steady mean luminosity , @xmath344 .
if the survey volume contains a total of @xmath325 smbhbs with the luminosity @xmath304 , then the periodic variable fraction , @xmath345 , can be determined down to the smallest value @xmath346 ( i.e. to find at least one periodic source ) . fixing the values of @xmath347 and @xmath348 ( as well as @xmath341 , @xmath24 and @xmath1 ) , this corresponds to a minimum variability time
scale @xmath349 that can be probed .
let us define the requirement that this minimum is @xmath350 weeks .
assuming that the longest variability time scale of interest is around @xmath351 year ( so that the periodic nature of the variations can be convincingly demonstrated over a multi year survey ) , this will offer a factor of three range in @xmath322 for mapping out the @xmath338 _ vs. _ @xmath352 dependence .
for example , with the steepest possible ( pure gw driven ) scaling @xmath339 , a survey volume containing a single source with @xmath353 weeks would then contain @xmath354 sources with a similar luminosity but with a @xmath355 week period . to fix some numbers ,
let us set @xmath356 , @xmath357 , and @xmath358 yr . for reference , the eddington luminosity of a @xmath359 bh at @xmath31 , assuming a @xmath192 10% bolometric correction , corresponds to an optical magnitude of @xmath29924 mag ( in the @xmath360 band ) .
let us also impose the ( somewhat ad hoc ) requirement that the survey volume should contain at least @xmath361 sources with a detectable flux variations at the period of @xmath362 weeks . in the gw driven stage , there will then be at least 5 detectable periodic sources with a period of @xmath363 weeks ; in the gas driven regime , where the scaling @xmath25 _ vs. _ @xmath322 is flatter , there will be a larger number of @xmath364week period sources . in figure
[ fig : survey ] , the curves show the sky coverage required to satisfy these criteria , as a function of the @xmath360band variable magnitude corresponding to the detection limit of the survey .
the bh masses producing the corresponding steady @xmath360 magnitude ( which , in our fiducial model , is 2.5mag brighter than the variable magnitude ) are shown on the top axis .
this figure assumes @xmath220 .
we used the fitting formula by @xcite for the bolometric quasar luminosity function ( lf ) @xmath365 to compute the the total number @xmath325 of quasars at @xmath31 , per solid angle @xmath342 , in a redshift range of @xmath366 , i.e. @xmath367 , where @xmath368 is the cosmological volume element , and @xmath369 is the bolometric luminosity corresponding to the steady magnitude threshold @xmath360 .
we then used equation ( [ eq : nvar ] ) for @xmath25 to obtain the total number @xmath370 of variable sources at observed period of @xmath353 weeks and at @xmath355 weeks . requiring @xmath338(@xmath355 weeks)@xmath371
then yields the solid angle @xmath342 as a function of @xmath360 .
note that the quasar lf is almost a pure power
law up to @xmath372 .
the break between @xmath360 = 26 - 27 mag in the solid curves corresponds to the transition between gw and gas driven orbital decay . in particular
, the figure shows that smbhs with a mass above / below @xmath373 are in the gw / gas driven regime , respectively .
figure [ fig : survey ] shows that there is a clear trade off between survey depth and area : the required sky coverage scales with the survey flux limit approximately as @xmath374 , with a steepening for shallow surveys with limiting magnitudes @xmath375 ( due to the decline at the bright end of the quasar lf ) , and a flattening for very deep surveys with limiting magnitudes @xmath376 ( because the smbhbs are in the gas driven regime and their residence times at fixed @xmath16 are shorter than in the pure gw driven regime ) . from figure
[ fig : survey ] , we conclude that , for example , a 1 sq . degree survey , detecting smbhbs whose steady luminosity is @xmath377 mag , with a variability at the level of @xmath378 mag , with sufficient sampling and duration to cover periods of 2060 weeks , represents an example for the minimum specification for the survey parameters ( i)-(iv ) . in this example , the mass of the bhs being detected is @xmath379 .
figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] show that at orbital periods of @xmath380 weeks , these bhs are all in the gw driven regime when @xmath235 , but may be in the viscosity driven regime for @xmath159 .
surveys that go deeper and cover a smaller area will begin probing the gas driven evolutionary stages . in figure
[ fig : survey2 ] , we examine how the required sky coverage changes when the parameters @xmath1 , @xmath343 or @xmath318 are modified .
the middle ( black ) curve shows the sky coverage required to find 20 sources at @xmath381 weeks ( intermediate between the red and black curves in figure [ fig : survey ] ) , with our fiducial parameters , @xmath220 , @xmath357 , @xmath382 .
the green curve corresponds to changing the mass ratio to @xmath201 ; this increases / decreases the residence time in the gw / gas driven regimes relative to the @xmath220 case ( compare figs .
[ fig : tres_q1 ] and [ fig : tres_q0.01 ] ) , and therefore reduces / increases the required solid angle coverage .
the top pair of ( blue ) curves in figure [ fig : survey2 ] show variations when either @xmath343 or @xmath318 is decreased by a factor of 10 ( to 0.03 or 0.01 , upper and lower of the pair , respectively ) .
note that the survey volume requirement is more sensitive to @xmath343 ( whereas the critical bh mass is equally sensitive to either ) .
similarly , the bottom pair of ( red ) curves show variations when either @xmath343 or @xmath318 is increased by a factor of 10 .
the small break visible at @xmath383 mag in this case corresponds to the transition from the middle to the outer disk region for @xmath384 smbhbs at @xmath385 weeks ( see fig . [
fig : tres_q1 ] ) .
each dashed curve shows the mass of the smbhb corresponding to the effective @xmath360 magnitude limit ( bh masses are labeled on the right @xmath386-axis ) .
the required survey volume also shifts linearly with the assumed total quasar lifetime @xmath317 . if we knew where to look ( i.e. , if _ lisa _ delivers a candidate for an on going merger , with sufficiently accurate localization on the sky ) , it would be possible to perform a deep , targeted observation for variability on short time
scales ; between several minutes up to @xmath387 hours within the last @xmath192 month of merger ( this possibility is discussed in detail in * ? ? ? * ) . however , each source will spend only a @xmath192month at such short variability time
scales , and a random search , in the absence of a preferred direction on the sky , would then have to monitor @xmath388 agn to find a single example of such a late stage periodic source .
alternatively , one may monitor @xmath389 agn for @xmath12 years , to look for ( slowly evolving ) periods , on the timescale of @xmath192 a day .
the slow decrease in the period , which would be a smoking gun for gw inspiral , will be challenging to observe in real time for individual objects .
existing observations from radio to x ray bands have shown that the luminosity of quasars and other active galactic nuclei varies on time scales from hours to several years ( see , e.g. , the articles in @xcite or the recent review by @xcite ) .
in fact , variability often aids in the identification of agn ( and may conversely be a major obstacle in identifying the periodic signal proposed here ) .
while variability is detected in a large fraction of all agn , there are only a handful of sources whose structure function shows clear _ periodic _ variability on long ( @xmath390 weeks ) time scales ( see , e.g. @xcite for a review focusing on searches for periodic variability ) .
examples include a handful of blazars , whose historical light
curves show periodic outbursts on timescales of a year to a decade , or even longer ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references in these papers , for individual objects ) .
@xcite monitored 25 low redshift , optically selected quasars for variability over several decades in infrared bands .
they identified one object , the radio loud quasar pg 1535 + 547 , whose structure function shows a periodic component with a period of @xmath387 yr .
this source has a bolometric luminosity of @xmath391 , implying a bh mass of @xmath392 .
figure [ fig : tres_q1 ] shows that equal
mass smbh binaries with this mass are in the gw driven regime , and @xmath393 may exhibit a 10year period . in comparison ,
figure [ fig : tres_q0.01 ] shows that unequal mass binaries with this total mass may be in the gas driven regime , and periodic variability may be exceedingly rare . thus , we conclude that the identification of one periodic object is roughly consistent with it being an example of a gw driven , near equal mass binary . however , there are only 4 objects in the sample studied by @xcite with luminosities above @xmath394 , prohibiting robust conclusions
. very large area variability surveys , such as in the sloan digital sky survey ( sdss ) , are shallow , and detect variability down to only @xmath395 mag @xcite . with our fiducial @xmath396
, this corresponds to periodic agn whose steady luminosities are @xmath397 mag .
figure [ fig : survey ] shows that such a survey would have missed the periodic variations discussed above .
figure [ fig : survey2 ] shows that if the variable fraction is much larger , @xmath398 , then @xmath399 variable sources with a yearly period would be detectable but their periodic nature could be demonstrated only with sufficient time
sampling , extending over a decade .
a recent , much deeper optical survey by the subaru telescope @xcite for variable objects provides interesting constraints on the scenario envisioned here .
the completeness function in this survey , defined as the probability to detect flux variations of an object with a variable component @xmath360 , goes from @xmath192unity to @xmath192zero between @xmath400 to @xmath401 ( see fig . 8 in @xcite ) . in the fiducial case with @xmath382 and @xmath357
, the limiting variability magnitude @xmath402 mag corresponds to the mean steady magnitude of @xmath403 mag , and bh mass of @xmath404 . at @xmath31 ,
the subaru survey has a completeness of 0.5 at this magnitude ( see fig .
11 in @xcite ) and covers an area of 0.9 sq .
this combination of sensitivity and area lies very close ( just below ) the curves in figure [ fig : survey ] . using the @xcite quasar lf , and assuming a completeness of 0.5
, we find that the subaru survey should detect @xmath405 agn ; this is in nearly exact agreement with their quoted result ( 489 deg@xmath406 ) .
we further find that of these sources , @xmath407 would vary with observed periods of 20 weeks , 60 weeks , and 1000 days ( adopting a probability of 0.5 for detecting variability , from fig . 8 in @xcite ) .
figure 12 in @xcite shows that they found several dozen sources that varied , at least once in their life , on all of these timescales .
unfortunately , we do not know whether these sources are periodic or not , and therefore the subaru survey results represent only an upper limit on the fraction of periodic sources .
nevertheless , this already suggests that the @xmath408 bhs at the limit of the survey can not produce variability at the level significantly exceeding our fiducial @xmath409 . the ultra deep
_ hubble space telescope ( hst ) _ variability surveys ( see , e.g. , the recent review by * ? ? ?
* and references therein ) discovered galaxies whose nuclei varied by magnitudes down to @xmath410 . the observations were taken a year apart in the hubble deep field north ( hdfn ) and the groth survey strip ( gss ) , whose areas are @xmath411 and @xmath412 sq .
degrees , respectively .
while the solid angle of the hdfn dataset is too small to yield useful constraints , the fiducial case with @xmath356 and @xmath357 in figure [ fig : survey ] shows that the gss dataset just reaches the sensitivity / area combination of @xmath413 mag and @xmath412 sq .
degrees required to find flux variations from @xmath414 smbhbs .
approximately @xmath415 of agn were found to vary by magnitudes down to @xmath416 in this dataset ( see * ? ? ? * for more details ) , suggesting that @xmath417 of agn containing smbhbs with this mass can produce variability at the @xmath418 level .
agn are also known to vary on long times
scales in x - ray bands .
systematic and unbiased variability surveys sensitive to times scales of weeks , such as those in soft x rays in the rosat all sky survey @xcite or in hard x rays in _
swift_/bat data @xcite however , have been restricted to the brightest agn , while deeper surveys , such as those by _ rxte _ @xcite , of the _ chandra _ deep field north @xcite and south @xcite , and by _ xmm _ @xcite have only monitored up to a few hundred sources .
these observations do suggest that a large fraction of agns vary in x ray bands on time scales of a day to a year , but whether the variations are periodic have not been determined . in the 9month duration observations covered by _
swift_/bat data , @xcite find a strong anti
correlation between luminosity and variability ( with no source with luminosity @xmath419 showing significant variability ) ; @xcite report a similar trend from an x - ray variability analysis of 66 agn in the lockman hole .
these findings would be consistent with the trend that the most massive smbhbs ( @xmath420 ) spend less time at a fixed orbital timescale of @xmath421 weeks ( see figures [ fig : tres_q1 ] and [ fig : tres_q0.01 ] ) .
the results of @xcite suggest that absorbed sources vary more than unabsorbed ones , which may be particularly relevant for finding the periodic smbh binary sources envisioned here , which are undergoing the last stages of their merger , and may be heavily obscured and visible primarily in x ray bands .
while the deep existing optical surveys come close to placing useful constraints on the scenario envisioned here , future surveys , designed to uncover source populations with periodic variations on times scales of tens of weeks , should be able to either discover these populations , or place stringent limits on their existence .
many large optical / ir surveys are being planned or built , motivated largely by finding type ia supernovae ( sne ) for cosmological studies ( see , e.g. , * ? ? ? * for a recent review ) .
the most ambitious of these , such as lsst and pan - starrs-4 will be all sky surveys , and should be able detect variability to @xmath422mag , allowing detections well beyond the most pessimistic case shown in figure [ fig : survey2 ] .
the proposed alpaca survey @xcite will cover 1,000 sq .
degrees to 23 - 25 mag in 5 optical bands , and would already reach the sensitivity / area combination probing these pessimistic scenarios .
in addition to producing periodic variability , there could be several other methods to prove or disprove the presence of a smbh binary .
first , the orbital motion of the binary may cause relative shifts in the quasar s emission lines .
for example , in a configuration in which the broad lines arise from gas close to one of the two ( moving ) bhs , and the narrow lines arise from material farther away , which is close to rest at the systemic redshift , such a shift could arise between the narrow and broad emission lines @xcite . similarly ,
if both bhs carry their own accretion disks , extending to a few schwarzschild radii , and produce broad lines @xcite , then there could be two sets of broad emission lines , super
imposed with a similar relative velocity shift .
the magnitude of these shifts may be of order the orbital velocity ( @xmath423 km / s at @xmath424 schwarzschild radii ) , which could be detectable either in individual objects , or else statistically for the population .
@xcite recently reported a candidate smbh binary , with two sets of broad emission lines separated by @xmath425 .
the spectrum of this source can also be interpreted with a single bh+disk system @xcite ; indeed , this interpretation is favored by the lack of any change in the velocity offset over the course of @xmath426 year @xcite .
nevertheless , it is interesting to note that , with the binary parameters reported for this source ( assuming random orientation , and an expected orbital speed of @xmath427 ) , @xmath428 , @xmath429 , @xmath430 , @xmath431 years , and @xmath432 , we find the evolutionary track of the proposed system to be virtually indistinguishable from the @xmath200 , @xmath201 case shown in figure [ fig : tres_r_q0.01 ] . at its currently observed orbital separation of @xmath432 , the binary would be in the gas driven regime , close to outer radius of the formally gravitationally stable disk ( i.e. , the system is just outside the marked @xmath256 point in this figure ) , with a residence time of @xmath433 years .
this could indeed make this observed separation common among quasars with @xmath200 smbhbs .
the figure also shows that the residence time at fixed orbital velocity decreases steeply with bh mass , suggesting that fainter binary quasars with similar orbital speeds would be much less common . in the last stages of coalescence , the gws emitted by such a system would induce periodic modulations in the arrival times of pulses from background radio pulsars ; at 200ns timing sensitivity
, these modulations would be detectable from smbhbs out to a distance of @xmath434mpc ( * ? ? ?
* note that this study already applies the idea to the source 3c66b mentioned in 1 ,
whose elliptical motion was interpreted as due to a smbhbs , and rules out the smbhb hypothesis ) .
in this paper , we followed the evolution of smbh binaries , starting from large separations , to coalescence .
we find , in agreement with earlier works , that the orbital decay is initially generically driven by viscous binary disk interactions , whereas gws dominate the last stages . in a refinement of earlier results
, we also find that just prior to the transition to gw driven evolution , the viscous orbital decay is generically in the `` secondary dominated '' type ii migration regime ( the mass of the secondary is larger than the enclosed disk mass ) .
this is slower than the disk dominated type ii migration that has sometimes been assumed in the past , and , as a result , smbh binaries spend a significant fraction of their time at orbital periods of @xmath192days to @xmath192 a year , where they may not be rare , and may be identifiable .
we emphasized the large uncertainties in the residence times in this regime for example , time
dependent disk models predict even slower decay .
we also find that observations of bhs with a mass range of @xmath435 over this range of periods could find binaries located in all three physically distinct regions of the circumbinary disk .
thus , several aspect of disk physics could potentially also be probed in future observations of a population of smbh binaries .
we also find that viscous processes may contribute to the orbital decay rate even after the binaries enter _ lisa _ s frequency range , for low and/or unequal mass binaries ( @xmath436 or @xmath437 ) . while viscous processes are strongly sub
dominant for rapidly evolving `` inspiral '' sources , detected during the last few years of their coalescence , the presence of the gaseous disk could reduce any background of unresolved stationary sources at frequencies near the low frequency end of the _ lisa _ range ( @xmath438 mhz ) .
we considered the possibility that there may be a one to
one correspondence between the activation of luminous agn and smbh coalescences , with a fraction of agn exhibiting periodic flux variations .
given that the interpretation of individual smbhb candidates have so far remained ambiguous , we proposed that a statistically large sample should aid in the identification of these binary bh sources .
our main conclusion is that future surveys in optical and x ray bands , which can be sensitive to periodic variations in the emission from @xmath439 supermassive black hole binaries , on timescales of @xmath5 tens of weeks , at the level of @xmath211% of the eddington luminosity , could look for a population of such sources , with the aim of determining the fraction @xmath25 of sources , at a given redshift and luminosity , as a function of @xmath322 . in our simplified models for the binary
disk interaction , this time
scale of tens of weeks corresponds to the orbital time when binaries with @xmath440 make their transition from viscous to gw driven evolution . in the latter regime , for sources with @xmath441 , gravitational radiation predicts the scaling @xmath442 .
the discovery of a population of periodic sources whose abundance obeys this scaling would confirm that the orbital decay is indeed driven by gws , and also that circumbinary gas is present at small orbital radii and is being perturbed by the bhs .
deviations from the @xmath443 power
law for lower mass bhs would constrain the structure of the circumbinary gas disk and viscosity driven orbital decay .
there is certainly a possibility that the periodic sources envisioned here do not exist ( e.g. , because the smbh binary does not produce bright and variable emission during its gw emitting stage , at orbital separations of @xmath444 schwarzschild radii ) .
nevertheless , we argued that existing surveys already approach the required combination of sky coverage and depth , and future surveys , designed to make observations for several years , with a sampling rate of a few days , could yield a positive detection and identify periodic source populations .
this would bring rich scientific rewards , possibly including the indirect detection of gravitational waves , driving the orbital decay of these sources .
zh thanks george djorgovski and tuck stebbins for stimulating discussions , and mamoru doi and tomoki morokuma for sharing their subaru variability search results prior to publication , which originally inspired this paper .
we also thank zsolt frei and david hogg for useful comments , chris stubbs , michael strauss and richard mushotzky for advice on variability surveys , and the anonymous referee for comments that significantly improved this paper .
km thanks the aspen center for physics , where a part of the work reported here was performed , for their hospitality .
this work was supported by the polnyi program of the hungarian national office for research and technology ( nkth ) and by nasa atfp grant nnx08ah35 g .
bk acknowledges support from otka grant 68228 . | supermassive black hole binaries ( smbhbs ) in galactic nuclei are thought to be a common by product of major galaxy mergers .
we use simple disk models for the circumbinary gas and for the binary - disk interaction to follow the orbital decay of smbhbs with a range of total masses ( @xmath0 ) and mass ratios ( @xmath1 ) , through physically distinct regions of the disk , until gravitational waves ( gws ) take over their evolution . prior to the gw
driven phase , the viscous decay is generically in the stalled `` secondary dominated '' regime .
smbhbs spend a non negligible fraction of a fiducial time of @xmath2 years at orbital periods between days @xmath3 year , and we argue that they may be sufficiently common to be detectable , provided they are luminous during these stages .
a dedicated optical or x ray survey could identify coalescing smbhbs statistically , as a population of periodically variable quasars , whose abundance obeys the scaling @xmath4 within a range of periods around @xmath5 tens of weeks .
smbhbs with @xmath6 , with @xmath7 , would probe the physics of viscous orbital decay , whereas the detection of a population of higher mass binaries , with @xmath8 , would confirm that their decay is driven by gws .
the lowest mass smbhbs ( @xmath9 ) enter the gw - driven regime at short orbital periods , when they are already in the frequency band of the _ laser interferometric space antenna _ ( _ lisa _ ) .
while viscous processes are negligible in the last few years of coalescence , they could reduce the amplitude of any unresolved background due to near stationary _ lisa _ sources .
we discuss modest constraints on the smbhb population already available from existing data , and the sensitivity and sky coverage requirements for a detection in future surveys .
smbhbs may also be identified from velocity shifts in their spectra ; we discuss the expected abundance of smbhbs as a function of their orbital velocity . |
ion - atom collisions have been an attractive subject and are of considerable interest in atomic physics due to both their importance in fundamental physics and their application in many fields , such as laboratory and astrophysical plasmas @xcite , heavy ion inertial fusion @xcite , radiation physics , collisional and radioactive processes in the earth s upper atmosphere @xcite and many other technological areas . in recent decades
, ion - atom collisions have been studied in detail experimentally as well as theoretically from low to relativistic collision energies ( see , for example , refs .
there has been an increased need the evaluation of ion - atom cross sections of different processes for many accelerator applications shevelko .
for example , a beam interaction with the remaining background gas and gas desorbed from walls limits the intensity of bunches at the rhic ( relativistic heavy ion collider ) @xcite and a pressure rise from ion losses at the low - energy antiproton ring brought concerns for the lhc ( large hadron collider ) @xcite . at moderate energies , collisions of closed - shell ions with closed - shell atoms for various inelastic channels such as the ionization , charge - exchange , stripping , and excitation
are well understood .
however , the absolute cross section information of these inelastic processes are of interest - in particular , the energy dependence of collisions of k@xmath7 ions with he atoms because the excitation probabilities in these collisions ( with some exceptions ) are one or two orders of magnitude smaller than in symmetric and quasisymmetric systems @xcite .
another remarkable feature of inelastic collisions of these closed - shell particles is that the excitation into a doubly excited state occurs with a high probability .
therefore , an accurate determination of the structure of different inelastic cross sections for these collisions is important to understand the mechanisms for inelastic transitions in the outer and , in some cases , inner shells of colliding atomic particles . in order to discuss quantitatively the excitation mechanism one has to evaluate the parameters of existing theories and explore the contributions of separate inelastic channels for the investigated processes by using experimental measurements . despite many studies of k@xmath0he collisions which have been carried out by various methods @xcite , available data for the absolute cross sections of the above mentioned processes are contradictory @xcite and , in some cases , unreliable @xcite .
charge exchange cross sections for k@xmath0ar , k@xmath0kr , k@xmath0xe and k@xmath8he systems were reported in ref .
@xcite using the detection of fast neutral particles within a definite interval of scattering angles .
however , as shown in ref .
@xcite the restriction on the interval for collision angles in ref .
@xcite underestimated the measured charge exchange cross sections by a factor of ten for the k@xmath0ar , kr , xe collision pairs over the entire energy range considered .
the charge exchange cross sections reported in ref .
@xcite might also have been underestimated for k@xmath7 - he collisions .
hence , it was necessary to carry out measurements of the charge - exchange cross sections for k@xmath0he collisions in a wider and more complete interval of scattering angles using a method that is free of this deficiency . in ref .
@xcite , only the absolute value of the cross section for excitation processes , occurring in the case when the mass of the incident ion is less than of target particle ( na@xmath0ar ) was reported .
the results of the measurements of the excitation function in arbitrary units for na@xmath8he and k@xmath0he collision pairs are presented in refs .
matveev , matveev2 .
the lack of absolute excitation cross sections for the asymmetric k@xmath0he system motivated the present detailed investigation of the primary mechanisms for this collision process . in order to make use of the measured cross section for electron production ,
it is necessary to know absolute ionization cross sections for the target atoms and absolute stripping cross sections for the incident ions . since there are no previous measurements of the stripping cross section for k@xmath9he collisions , it is imperative to measure this cross section .
as is known , the basic mechanism responsible for the process of ionization in atomic collisions is a diabatic orbital super promotion into the continuum .
proposed in ref .
@xcite , this mechanism was further developed in refs .
@xcite . advantages and limitations of these studies are assessed for colliding systems with one active electron , e.g. h@xmath0h and he@xmath10h .
it is important to extend the theory to ionization processes in collisions of many
electron atomic systems .
experimental methods @xcite that are used so far to study this mechanism are based on the measurement of continuous parts of the energy spectra of ejected electrons .
more detailed information for the ejected electrons at different impact parameters can be obtained using coincidence techniques .
one of the objectives of this work is to show that in many cases the information extracted from complicated coincidence experiments can also be obtained using a simple method , namely by measuring the energy loss spectra of incident particles .
such spectra can be measured over a wide range of scattering angles and energy losses , for which the problems of collection of low energy electrons do not arise .
below , we report absolute total cross sections for charge exchange processes , ionization , and stripping processes that result in the production of free electrons , as well as the excitation of both the projectile and target particles and energy loss spectra in collisions of k@xmath7 ions with he atoms .
the remainder of this paper is organized in the following way . in sec .
exper we introduce our unique experimental setup that includes three collision chambers and present the experimental techniques and procedures used to measure the absolute total cross sections for the excitation , charge exchange , ionization , and stripping processes .
results of measurements and discussion of mechanisms for different processes occurring in k@xmath0he collisions are given in sec .
[ results ] . finally , in sec .
[ concl ] we summarize our studies and present the conclusions .
a schematic view of the experimental setup is shown in fig
. 1 . a beam of k@xmath11 ions from a surface
ionization ion source is accelerated and focused by an ion - optics system , which includes quadruple lenses and collimating slits .
after the beam passes through a magnetic mass spectrometer , it enters the first collision chamber containing he gas for measurements of excitation processes in k@xmath0he collisions .
our setup is designed so that when the beam of k@xmath7 ions passes through the second collision chamber measurements of the charge exchange , ionization and stripping processes occur .
the energy loss spectrum is obtained by a collision spectroscopy method when k@xmath7 ions collide with target he atoms in the third collision chamber .
the pressure in each collision chamber when there is no he target gas is kept below 10@xmath12 torr and the typical pressure under operation is 10@xmath13 torr .
the measurements are performed under a single - collision conditions .
the current of primary ions in the collision chamber is @xmath14=0.01 - 0.1 @xmath15a .
the uniqueness of our experimental approach is that the quality of the beam as well as the experimental conditions for all processes under investigation always remain identical .
moreover , the experimental techniques include a condenser plate method , angle- and energy - dependent collection of product ions , energy - loss and optical spectroscopy under the same `` umbrella '' .
the measurement procedures have been discussed in some detail previously @xcite , therefore the description of the present measurements can be kept comparatively short .
the measurements of the excitation processes in k@xmath0he collision were performed when the beam of k@xmath7 ions passed through the first collision chamber .
cross sections for the excitation processes were measured by the optical spectroscopy method @xcite .
the radiation emitted in the first collision chamber was observed at an angle of 90@xmath16 with respect to the direction of the primary ion beam .
the spectral analysis of this radiation was performed in the vacuum ultraviolet ( vuv ) as well as visible spectral regions .
an electron multiplier was used to detect the intensity of the radiation .
particular attention was devoted to the reliable determination and control of the relative and absolute spectral sensitivity of the light - recording system .
this was done by measuring the signal due to the emission of molecular bands and atomic lines excited by electrons in collisions with h@xmath17 , n@xmath17 , o@xmath17 and ar . for this , an electron gun was placed directly in front of the entrance slit of the collision chamber . the relative spectral sensitivity , and the values of the absolute cross sections ,
were obtained by comparing the cross sections for the same lines and molecular bands reported in the literature @xcite .
the uncertainties in the excitation cross sections for the k @xmath8he system are estimated to be 20% and the uncertainty of the relative measurements does not exceed 5% .
when the k@xmath7 beam enters the second collision chamber filled with the target gas , the measurements for the charge exchange and ionization processes occur .
the charge exchange and ionization cross sections were measured by a refined version of the capacitor method @xcite . in an earlier paper @xcite the measurements
were performed by the standard transfer electric field method .
the customary procedure is to use one of the central electrodes as the measurement electrode .
we consider that such approach is the reason for significant errors in measurements ogurtsov because scattered k@xmath7 ions may strike the electrodes used for measurements .
this effect becomes more evident especially for low energy collisions . to avoid this deficiency we accordingly used a refined version of the transfer electric field method in which the effect of scattering on the measured results is substantially reduced by shifting from the central electrode ( standard method ) to the first electrode ( towards beam entrance side ) . due to fringing effects at the edges of this electrode
a system of auxiliary electrodes between the first electrode and the entrance slit was installed .
these auxiliary electrodes create a uniform potential near the first electrode .
the first electrode , the auxiliary electrodes , and the entrance slit are all positioned as close together as possible .
this close arrangement limits the scattering region on the beam entrance side .
thus , according to our estimation , taking into account the geometry of our facilities , only those ions which are scattered through angles greater than 70@xmath16 can reach the electrodes .
the primary ions are detected by the faraday cup .
collision particles ( secondary positive ions and free electrons ) are detected by a collector .
the collector consists of two rows of plate electrodes that run parallel to the primary ion beam . a uniform transverse electric field , responsible for the extraction and collection of collision particles , is created by the potentials applied to the grids .
this method yields direct measurements of the cross sections @xmath18 for the production of singly positively charged ions and @xmath19 for electrons as the primary beam passes through the gas under study .
these measured quantities are related in an obvious way to the capture cross section @xmath20 and the apparent ionization cross section @xmath21 , and are determined as @xmath22 in @xmath23 is the cross section stripping of the incident ion .
the cross section @xmath21 is always larger than the cross section for stripping @xmath24 it should be mentioned that if the cross sections for multiple ionization and ionization with capture are small then @xmath25 . the uncertainty in the measurements of the absolute values of the cross sections @xmath18 and
@xmath19 is estimated to be 15% over the entire energy interval studied .
this is determined primarily by the uncertainty in the measurement of target gas pressure in the collision chamber .
the uncertainty in the measurement of the ionization cross section @xmath26 is estimated to be @xmath27 over the entire energy range . the larger value in comparison with that for the measurement of the cross section
@xmath19 is a consequence of the presence of an additional error in the measurements of @xmath21 due to stripping of k@xmath7 ions . at k@xmath7 energies less than 2.0
kev the cross section @xmath18 is significantly larger than @xmath19 .
accordingly , the error in the determination of the capture cross section @xmath20 in this energy region is determined primarily by the error in the measurement of @xmath28 .
with increasing k@xmath7 energy the cross sections @xmath18 and @xmath29 become more nearly equal . as a result the error in the determination of @xmath20 increases . at a k@xmath7 energy of 5 kev
it is estimated to be 25% .
the stripping cross sections were measured in an independent experiment . a beam of k@xmath7 ions from a surface - ionization ion source after acceleration and focusing passes through the second collision chamber .
it is then analyzed for the charge composition by an additional 90@xmath16 magnetic mass spectrometer ( not shown in fig.1 ) .
a set of slits that were located in front of the detector enable the profile of the projectile beam to be to scanned in detail .
the ratio of the number of doubly charged ions , produced as a result of stripping , to the incident ion beam , k@xmath30k@xmath31 is determined from the areas under the corresponding lines in the mass spectrum . since inelastic processes , apparently including stripping , may be accompanied by the scattering of particles through comparatively large angles in collisions between alkali metal ions and inert gas atoms afrosimov1 , afrosimov2 , afrosimov3 , special measures were taken to arrange complete collection of k@xmath32 ions .
the energy loss spectrum was obtained by the collision spectroscopy method @xcite . the primary beam extracted from the ion source
was accelerated to the desired energy before being analyzed according to @xmath33 ( @xmath34 and @xmath35 are the ion s charge and mass , respectively ) .
the analyzed ion beam was then allowed to pass through the third collision chamber by appropriately adjusting the slits prior to entering into a ` box ' type electrostatic analyzer .
the energy resolution of this analyzer is @xmath36=1/500 .
automatic adjustments of analyzer potentials gave the possibility for investigation of energy loss spectra in the energy range of 0 100 ev .
the differential cross section is measured by rotating the analyzer around the center of collisions over an angular range between 0 and 25@xmath16 degrees .
the laboratory angle is determined with respect to the primary ion beam axis with an accuracy of 0.2@xmath37 for the measurements of the charge exchange differential cross section the charge component of scattered primary particles realized in the third collision chamber is separated by the electric field ( see ion charge deflector in fig . 1 ) and neutral particles formed by electron capture collisions are registered by the secondary electron multiplier .
such a tool gives us the possibility to determine total cross sections and compare them with the results obtained in the second collision chamber .
in addition , the measured energy loss spectrum gives detailed information related to the intensity of inelastic processes realized in the first collision chamber ( excitation ) and in the second collision chamber ( charge exchange , ionization and stripping ) .
the collision of a k@xmath7 ion beam with he gas leads mainly to the following processes : charge exchange reaction @xmath38 when the k atom and he@xmath7 ion products can be in the ground state or in different excited states ; different excitation processes @xmath39 that include different channels for excitation of the k@xmath7 ion or / and he atom ; stripping of k@xmath7 @xmath40 and ionization of he @xmath41 as well as excitation of autoionization states @xmath42 when he@xmath43 is formed in different autoionization states and then decays by emitting an electron . in figs .
2 4 the measured energy dependences of absolute total cross - sections are presented for charge exchange into the ground and excited states of k@xmath7 , the ionization , stripping and excitation function of potassium atom , ion and target helium atom in k@xmath0 he collisions .
5b @xmath44 5d represent a typical example of the energy loss spectrum for k@xmath9he system . for comparison , in fig .
5a is given the energy loss spectrum for k@xmath0ar collision .
we have also measured the energies of the electrons emitted in the collisions .
it was found that the energy of most liberated electrons is below 10 - 15 ev . as seen from figs . 2 and 3
, a distinctive feature of most of these inelastic processes is the small magnitude of the cross section .
the excitation functions of the k ii lines at @xmath45 nm and @xmath3 nm are similar ( curve 1 and 5 in fig .
2 ) . the curves did not differ in shape and curve 1 could be obtained from curve 5 by multiplying the ordinate of the last one by a factor of @xmath46 5 .
another feature that is prominent in the ionization , charge exchange and excitation cross sections is the difference in the energy dependences . while the ionization cross section ( fig .
3 ) and excitation function of the k@xmath7 ion ( fig .
2 , curve 5 ) and k atom ( fig . 2 curve 2 ) are small at low energies and increase monotonically with energy up to 5 kev , the total charge exchange cross section has a complex energy dependence and generally decreases with the collision energy up to 5 kev and then increases ( fig .
2 , curve 3 ) .
structural features are also observed in the resonance line for the helium atom ( fig .
2 , curve 4 ) .
distinct features are observed in the energy loss spectra shown in figs .
5 . in the case
for which colliding particles have nearly equal masses ( k@xmath0ar , fig . 5 ( a ) ) the main inelastic process
is realized effectively in the 10 @xmath44 35 ev energy loss interval , while for asymmetric colliding pairs ( k@xmath8he , fig .
5d ) this energy losses are extended up to 60 ev .
the data obtained in this study can be used to draw certain conclusions about the possible reasons for these features in the cross sections and mechanisms for the corresponding processes . to explain these mechanisms we use a schematic correlation diagram of the adiabatic quasimolecular terms for the system of colliding particles .
the diagram was constructed based on barat - lichten rules @xcite and is presented in fig .
these results motivate us to explore the excitation functions for collisions of k@xmath7 ions with helium atoms . the radiation generated in these collisions
is shown in fig .
2 and it includes : one resonance line of the k atom at @xmath47 nm ( curve 2 ) corresponding to the channel k@xmath7(3p@xmath48 ) @xmath49 he(1s@xmath50 ) @xmath51k(4s)@xmath49he@xmath7(1s ) of process ( ce ) , two resonance lines of the k@xmath7 ion @xmath52at @xmath45 nm ( curve 1 ) and @xmath3 nm ( curve 5 ) for process ( [ k excited ] ) and one resonance line of the he atom at @xmath53 nm ( curve 4 ) corresponding to different channels of process ( [ k excited ] ) .
the kii @xmath45 nm line representing a transition from the @xmath4p@xmath54(3s@xmath503p@xmath55 [ @xmath50p@xmath56@xmath164s@xmath57 ) level to the ground state @xmath4s@xmath58(3s@xmath503p@xmath48 ) is much stronger than the @xmath59 nm line , corresponding to transitions from the state @xmath6p@xmath54(3s@xmath503p@xmath55[@xmath50p@xmath60@xmath164s ) to the ground state .
the intensity ratio of the k ii @xmath45 nm and @xmath3 lines shown in fig . 2 is 5:1 , indicating that the probability of excitation of the singlet resonance level @xmath4p@xmath5 is greater than that of the triplet level @xmath6p@xmath5 , due to spin conservation .
curve 4 in fig . 2 for the excitation function of the he(1s2p ) resonance line for the channel k@xmath7(3p@xmath48 )
@xmath49 he(1s@xmath50 )
@xmath51k@xmath7(4p)@xmath61he(1s2p ) has unusual energy dependence .
this results from the overlapping of different excitation channels in different energy ranges . particularly , the analyses of the correlation diagram in fig .
6 leads to the conclusion that at relatively small energies the excitation of he(1s2p ) originate due to the exchange interaction through the intermediate state corresponding to the channel k@xmath7(3p@xmath48 ) @xmath49 he(1s@xmath50)@xmath51k(4s ) @xmath49he@xmath7(1s ) , while at large energies the excitation of he(1s2p ) may occur due to the @xmath62 transition between terms which correspond to the k@xmath7(3p@xmath48 ) @xmath49 he(1s@xmath50)@xmath51k@xmath7(4s@xmath57)@xmath49he(1s@xmath63 ) and k@xmath7(3p@xmath48 ) @xmath49 he(1s@xmath50)@xmath51k@xmath7(3p@xmath48)@xmath61he(1s2p ) channels . in determining the processes responsible for charge exchange in k@xmath0he collisions
, we compare the total charge exchange cross sections presented in fig . 2 by curve 3 with those corresponding to the decay of resonant levels of the potassium atom presented by curve 2 . taking into account the selection rules and the ratio of oscillator strengths for the transitions , we show that the decay of any of the excitation levels of the potassium atom culminates in about half the cases with a transition of the atom to a resonant state , which then decays . according to our data , the decay cross section of the resonant levels of the potassium atom in k@xmath7- he collisions increases with incident ion energy , reaching about @xmath64 cm@xmath50 at an ion energy of 3 kev and 7@xmath6610@xmath67 @xmath68 at 5 - 7 kev . at low collision energies
these cross sections are less than 1% , and at high collision energies are @xmath69 of the total charge exchange cross sections .
it follows that in these collisions the cross section for the capture of an electron to an excited state of atom through the channels k@xmath7(3p@xmath48 ) @xmath49 he(1s@xmath50 ) @xmath51 k(4p)@xmath49he@xmath11(1s ) or k@xmath7(3p@xmath48 ) @xmath49 he(1s@xmath50)@xmath51k(3d)@xmath49he@xmath7(1s ) is less than 10% of the total charge exchange cross section . in other words , the charge exchange process results primarily from capture of an electron to the ground state of the k atom .
the target ions are produced in these processes primarily in the ground state .
processes involving capture accompanied by excitation of the resultant target ion he@xmath7 make a small contribution into the charge exchange cross section , apparently because the energy defect of these processes is of the same order of magnitude as the collision energy ( the total kinetic energy of the particles in the c.m
. system ) @xcite .
the energy defect of these processes is @xmath4661 ev , while the collision energy at an ion energy of 700 ev is only 65 ev in the c.m . system .
we thus can conclude that the charge exchange reaction ( [ ce ] ) is governed primarily by the capture of an electron in the ground state of the potassium atom , accompanied by the formation of a helium ion also in the ground state : k@xmath7(3p@xmath48 ) @xmath49 he(1s@xmath50 ) @xmath51k(4s ) @xmath49 he@xmath11(1s ) .
this process can occur , as can be seen from the diagram in fig .
6 , as a result of the direct pseudocrossing of the term corresponding to the state k(4s ) @xmath49 he@xmath7(1s ) with the ground state of the system . since the k ( 4s ) @xmath49 he@xmath7(1s ) state
has only @xmath70 symmetry , it follows that @xmath71 transitions play a dominant role in the charge exchange processes .
- ar collision .
( b)-(d ) energy loss spectra for k@xmath7- he collisions at different scattering angles.,title="fig : " ] - ar collision .
( b)-(d ) energy loss spectra for k@xmath7- he collisions at different scattering angles.,title="fig : " ] - ar collision .
( b)-(d ) energy loss spectra for k@xmath7- he collisions at different scattering angles.,title="fig : " ] - ar collision .
( b)-(d ) energy loss spectra for k@xmath7- he collisions at different scattering angles.,title="fig : " ] since the mass of projectile k@xmath7 is much larger than that of the he target , a significant decrease in the velocity of the relative motion of the particles when traversing the pseudocrossing region occurs at significantly large initial ion energies ( of the order of 1 kev ) , unlike for systems with comparable masses .
in this situation , the only particles which can approach each other to a distance at which the terms undergo a pseudocrossing are those for which the impact parameters are much smaller than this distance .
the small value of the charge exchange cross section in k@xmath0he collisions compared to those for collisions of certain other pairs of particles , e.g. , k@xmath0ar , k@xmath0kr , k@xmath0xe , at ion energies 0.7 - 10.0 kev , can be attributed precisely to this effect @xcite . for these pairs of particles , the mass of the incident particle is less than or roughly equal to the mass of the target particle . therefore , there is no significant decrease in the relative velocity compared to the initial velocity of the ion .
correspondingly , there is no significant decrease in the impact parameters at which the pseudocrossing regions are reached . in the two - channel approximation ,
the charge exchange cross section should increase with increasing of ion energy , approaching the maximum value @xmath72 0.5@xmath73 , where @xmath74 is the position of the pseudocrossing region @xcite . using a limiting value of 65 ev as the interaction energy of the particles in the pseudocrossing region , and the potential curve for the ground state of k@xmath0he system from ref .
nikulin , we find @xmath75 @xmath76 0.7
and correspondingly , @xmath77 0.7@xmath6610@xmath78 cm@xmath50 .
it can be seen from fig . 2 ( curve 3 ) , that the cross section actually increases ( up to @xmath79 cm@xmath50 ) with increasing of the ion energy only up to 1.5 kev . as
the energy further increases , the cross section decreases ; and it increases again only at energies greater than 5 kev .
this behavior of the cross section indicates that at the ion energy of 1.5 kev the charge exchange mechanism is not correlated exclusively with the interaction of the entrance and exit channels , since the interference of these channels with other channels becomes important .
in particular , the decrease of the cross section at ion energies [email protected] kev can be related to the interference of the exit channel with the channel asymptotically dissociating into to the state k@xmath7(3p@xmath554s@xmath80)@xmath49he(1s@xmath50 ) , as follows from fig .
an interaction of these channels is confirmed by the agreement of the position of the minimum in the energy dependence of the charge exchange cross section with the position of the maximum in the energy dependence of the cross section for the decay of the resonant level of the ion k@xmath7(3p@xmath814s@xmath80 ) , shown in fig .
2 . the position of the maximum value of the cross section 1.5@xmath6610@xmath82 cm@xmath50 ( see fig .
2 , curve 1 ) corresponds roughly to that of the dip in the charge exchange cross section .
it follows from the results of present measurements of emitted electrons in k@xmath0he collisions and , in part , from the data obtained in ref .
aizawa , where the electron spectrum was measured over the interval 5 - 24 ev , that the liberation of slow electrons ( with energies less than 10 - 15 ev ) is a characteristic of ionization ( [ ioniz ] ) in k@xmath0he collisions . in order to determine the channel and mechanism of ionization
, we estimate the contribution of several inelastic processes that result in emission of slow electrons .
the contribution of the direct ionization k@xmath7(3p@xmath48)@xmath49he(1s@xmath50)@xmath51k@xmath7(3p@xmath48)@xmath49he@xmath7(1s)@xmath49e , that in the quasimolecular model is linked to the transition of the adiabatic term into the continuum , is estimated following ref .
the expression given in ref .
@xcite is written in terms of the principal quantum number for a hydrogen - like atom .
therefore , we modify it slightly for the estimation of the cross section for the emission of electrons from multielectron atoms . then the final expression can be written as : @xmath83 where @xmath84 is the relative velocity of the particles in the nonadiabatic region , @xmath85 is the number of electrons within the united atom in the state with quantum numbers ( @xmath86 , @xmath87 is the binding energy of an electron in the nonadiabatic region , and im@xmath88 and re@xmath88 are the coordinates where the potential surfaces cross in the complex plane .
they are @xmath89 where @xmath90 is the effective charge of the nucleus of the united atom .
the last formula was derived by the same approach as the expression ( 26 ) in ref .
@xcite , except that the population of the initial orbital is taken into account , and the binding energy is introduced as a parameter .
analysis of the correlations of the molecular orbital in the ( khe)@xmath11 system shows that in the united atom limit , the 1s electrons of the he atom become 3d electrons of the sc@xmath7 ion . since
this result is of importance for evaluating the contribution of the direct ionization , we note that it agrees both with the barat- lichten correlation rules @xcite and the eichler@xmath44wille rules @xcite .
consequently , the value @xmath91 was chosen for the evaluation of the cross section .
the binding energy @xmath92 of the electrons in the nonadiabatic region was chosen to be equal to that of the 3d electrons of the sc@xmath7 ion .
the effective charge @xmath93 was determined by the interpolation of the data of ref .
hartree . for the 3d electrons of the sc@xmath7 ion
, the value @xmath94 was found .
estimates of the cross section for the direct ionization with these parameters show that at an ion energy of 2.5 kev , the contribution of the direct ionization to the total ionization cross section is less than 0.1% , while at 6.5 kev it is less than 5.5% i.e. , this contribution is insignificant over the entire energy range studied .
the double ionization of the he atom , and capture accompanied by ionization of the he atom evidently make a small contribution to the ionization cross section .
there are two reasons for this : the large energy defect for these processes , 79 and 74.6 ev , respectively , and the absence of pseudocrossings of the corresponding quasimolecular terms with the ground - state term , as follows from the diagram in fig.6 . as follows from the electron spectrum in ref .
@xcite , the resultant intensity of the discrete lines associated with capture to the autoionization states of the k atom and ultimately corresponding to the ionization of the he atom is several times lower than the that of the lines of the k@xmath7 ion corresponding to stripping processes . since the stripping cross section is less than 10% of that for ionization ( see fig .
3 ) we find that the capture to autoionization states of k atoms also plays no important role in the ionization .
consequently , by systematically evaluating the contributions of various inelastic processes to the ionization of the target atoms in k@xmath0he collisions , we find that the ionization may be caused primarily by the decay of quasimolecular autoionization states . these could be expected to be states with two excited electrons ( see the energy loss spectrum in figs . 5c and 5d )
, since such states undergo quasimolecular decay with a high probability , liberating mainly slow electrons with a continuous energy distribution @xcite . both of these circumstances agree with the conclusions which follow from an analysis of the correlation diagram of the system , and they add a few refinements to the ionization mechanism .
it can be seen from the diagram that the ground state of the system goes over in the limit of the united atom into the 3p@xmath483d@xmath50 ( @xmath4d@xmath17 ) singlet state of the sc@xmath7 ion . below this state ,
the sc@xmath7 ion has only three triplet states ( not shown in the diagram ) and one singlet state ( with the configuration 3p@xmath483d4s ) @xcite .
only the term , asymptotically dissociating into k ( 3p@xmath484s ) @xmath49 he@xmath7 , with the appropriate @xmath95 symmetry , crosses the ground state .
note that electron transitions occur primarily between terms of the same multiplicity .
we find that the autoionization results from the filling of primarily terms of @xmath96 and @xmath97 symmetry .
these terms correlate in the limit of separated atoms to the states : k@xmath98(4s)@xmath44he@xmath99(1s2s ) , k@xmath100(4s)@xmath44he@xmath99(1s2p ) , k@xmath101 ( 3p@xmath1023d4s)@xmath44he(1s@xmath50 ) and k@xmath7(3p@xmath48)@xmath44he@xmath43(2p@xmath50 ) ( fig .
all of these terms , as expected , correspond to two - electron excitations of the system . since the ground - state term has the symmetry @xmath95 , the terms are populated as a result of @xmath62 and @xmath103 transitions , associated with the rotation of the internuclear axis along the different nuclear trajectories . according to our estimates ,
the governing mechanism for the stripping process ( [ striping ] ) is the transition of adiabatic term into the continuum .
the stripping cross section determined by this mechanism was calculated using eqs .
( [ sigma ] ) and ( [ rr ] ) . as it follows from the diagram in fig.6 the 3p electrons of the k@xmath7 ion
are correlated with the 3p electrons of the sc@xmath7 ion .
consequently , in evaluating the cross section we selected @xmath104 , while we took @xmath90 to be the same as for the 3d electrons of sc@xmath7 ion in evaluating the ionization cross section .
the stripping cross section calculated with these parameters is consistent over the entire energy region ( fig .
3 , curve 1 ) with the experimental cross section ( fig .
3 , curve 2 ) to within less than 25% .
although this agreement is fortuitous to some extent , it does support the conclusion that this is the governing mechanism for the stripping : k@xmath7(3p@xmath48)@xmath49he(1s@xmath50)@xmath105k@xmath32(3p@xmath55 ) @xmath49he(1s@xmath50)@xmath106 . as shown in fig .
4 , the excitation functions of the k atom and k@xmath7 ion for 4p lines are similar , apparently indicating a common mechanism of population . for the interpretation of this result
it is expedient to use the correlation diagram of molecular states of the ( khe)@xmath7 system from fig .
we emphasize that the term corresponding to the k(3d ) state is responsible for both
the population of the potassium atom and the ion into the 4p state .
the terms corresponding to the excitation of k(4p ) as well as k@xmath7(4p ) do not have immediate crossing points with the entrance term k@xmath7(3p@xmath48)@xmath44he(1s@xmath50 ) .
however , both of these terms may be populated from that corresponding to k ( 3d ) excitation states by means of a double rotational transition @xmath103 between the entrance term and that corresponding to the k(3d)@xmath44he@xmath7(1s ) state .
the excitation state of the term corresponding to k(4p ) may be populated from the term corresponding to the k(3d ) , via the cascade transition , from level 3d to 4p . as for the population of the term k@xmath7(4p ) , because the term k@xmath11(4p)@xmath44he(1s@xmath50 ) and k(3d)@xmath44he@xmath7(1s ) at large internuclear distance are energetically closed ( the energy defect is 0.4 ev ) , the exchange interaction between these states may cause a population of the term corresponding to k@xmath7(4p ) , and hence the identical energy dependences of the cross sections .
the probability of population of the k(3d ) level of the potassium atom itself is very small , due to the small probability of the @xmath107 transition ( see fig .
perhaps this is the reason for the anomalous small excitation cross section of the potassium atom in k(4p ) state and for the k@xmath7(4p ) state .
this can be seen from fig .
2 , curve 2 and fig .
4 , curves 1 and 2 , respectively .
another conclusion that can be drawn from fig .
5 is that the energy loss spectra for both k@xmath0ar and k@xmath0he collisions have a discrete character . the energy loss spectrum for k@xmath0ar in fig .
5 ( a ) is presented for comparison and is chosen for the fixed energy ( @xmath108 2 kev ) and scattering angle ( @xmath109 = 3.5@xmath16 ) at which the inelastic channels are most populated .
the first peak of this spectrum corresponds to the elastic scattering of k@xmath7 ions , while the second one corresponds to the single electron excitation of the argon atom to ( 4s ) , ( 4p ) and ( 3d ) states with energy losses of @xmath110 ev .
the third one corresponds to the excitation of k@xmath7 ions into the ( 4s ) and ( 3d ) states with energy losses @xmath111 ev , and the fourth one corresponds to the single and double autoionization states of ar with energy losses @xmath112 ev .
the energy loss spectrum for k@xmath0 he colliding pairs is given at @xmath113 kev and for the various scattering angles of k@xmath7 ions : @xmath114 , fig .
5 ( b ) ; @xmath115 , fig . 5 ( c ) ; @xmath116 , fig . 5 ( d ) respectively .
this allows us to observe the dynamics of inelastic channels .
let us analyze the energy loss spectrum for the k@xmath0@xmath113@xmath116 scattering angle shown in fig . 5 ( d ) .
the first maximum at zero energy loss is attributed to the elastic scattering of k@xmath7 ions .
the second maximum at @xmath117 ev is ascribed to the excitation of the potassium ion into the ( 4s@xmath118 ) states and to the direct excitation of the helium atom into ( 2p ) state . with low probability of excitation ionization and stripping , with energy losses @xmath119 ev and @xmath120 ev
, respectively , are favored .
rather higher excitation probabilities are observed at the energy loss @xmath121 ev and they are ascribed to the two electron excitation of the helium atom into ( 2s@xmath122 and ( 2p@xmath50 ) states hicks , kita2 . the remarkable feature of the spectrum from figs .
5 ( b)@xmath445 ( d ) is the rapid decrease of the elastic channel intensity with scattering angle .
this fact indicates the increasing importance of inelastic channels . in spite of the fact that the energy loss spectrum has been measured up to 60 ev
, a significant elastic peak was observed only at 1@xmath16 scattering angle . with the increase of scattering angle from 1@xmath16 to 3.5@xmath123 the excitation of the inelastic channel ( one and two electron excitation of helium atom and excitation of potassium ion ) becomes predominant .
we have also observed these specific features of the energy loss spectrum for other energies of the incident ions .
this allows us to conclude that for k@xmath0he collisions in this energy range , the main inelastic channel is the excitation of the k@xmath11 ion into the ( 4s@xmath57 ) state .
this result was also confirmed from our excitation function given in fig.2 , curve 1 .
indeed , as seen from this figure , the intensities of excited k@xmath7(4s@xmath124 ) lines are more than one order higher than for the rest of the inelastic channels considered here .
using refined experimental methods that include a condenser plate method , angle- and energy - dependent collection of product ions , energy - loss and optical spectroscopy under the same `` umbrella '' and a well checked calibration procedure of the light recording system , we have measured the absolute values of the charge exchange , ionization , stripping and excitation cross sections .
the correlation diagram of the ( khe)@xmath7 system has been employed to discuss the mechanisms for these processes in the k@xmath7 ion energy range @xmath125 kev .
charge exchange in k@xmath0he collisions mostly occurs through the channel k@xmath11(3p@xmath48 ) @xmath49 he(1s@xmath50 ) @xmath51 k(4s)@xmath49he@xmath7(1s ) resulting from the capture of the electron to the ground state of the atom in regions of pseudocrossings of the potential curves of @xmath95 symmetry .
the primary ionization mechanism in k@xmath0he collisions is the filling , as a result of @xmath62 and @xmath103 transitions , of quasimolecular autoionization terms and their decay in the region of the transition into the continuum ( in the stage in which a quasimolecule exists ) .
stripping in k@xmath0he collisions occurs via a mechanism involving a transition of adiabatic term into the continuum in the region of a non adiabatic interaction of molecular orbitals with orbital angular moments which are identical in the limit of the united atom . a common mechanism of the populations , as well as the small values of excitation cross sections for the potassium ion and atom into 4p states are explained .
the usage of the collision spectroscopy method has permitted us to observe exceptionally highly excited states of helium atom in the energy loss spectrum . | absolute cross sections for charge exchange , ionization , stripping and excitation in k@xmath0he collisions were measured in the ion energy range @xmath1 kev .
the experimental data and the schematic correlation diagrams are used to analyze and determine the mechanisms for these processes .
the increase of the excitation probability of inelastic channels with the angle of scattering is revealed .
an exceptionally highly excited state of he is observed and a peculiarity for the excitation function of the resonance line is explained . the intensity ratio for the excitation of the k ii @xmath2 nm and @xmath3 nm lines is 5:1 which indicates the high probability for excitation of the singlet resonance level @xmath4p@xmath5 compared to the triplet level @xmath6p@xmath5 . the similarity of the population of the 4p state of the potassium ion and atom as well as the anomalously small values of the excitation cross sections are explained . |
the need to maintain the thermodynamic stability of the native state is an important determinant of protein evolution .
this requirement has different effects on the various positions of the protein , depending on their structural environment .
therefore it is crucial to consider the effect of site - specific structural constraints in models of protein evolution , such as those used to estimate phylogenetic distances from the comparison of protein sequences ( nei and kumar , 2000 ) or to reconstruct phylogenetic trees through maximum likelihood methods ( felsenstein , 1981 ) .
the first and simplest models of this kind assumes a gamma distribution of site - specific substitution rates ( see nei and kumar , 2000 ) .
the gamma distribution is flexible enough to interpolate between broad and narrow distributions , and its free parameter improves considerably the fit between models of evolution and multiple sequence alignments .
subsequently , site - specific amino acid frequencies ( halpern and bruno , 1998 ) and site - specific substitution matrices determined for different structural classes ( li and goldman , 1998 ; koshi and goldstein , 1998 ; koshi et al . , 1999 ; thorne , 2000 ; parisi and echave , 2001 ; fornasari et al . , 2002 ) were shown to further improve the reconstruction of phylogenetic trees . in these works ,
site - specificity is either empirically derived from the data or simulated with the help of some protein model . following the latter approach
, we recently showed that a structural property , the principal eigenvector ( pe ) of the contact matrix , is a strong predictor of site - specific amino acid conservation in models of protein evolution with stability constraints ( bastolla et al . , 2004b ) .
we represent the amino acid sequence through a real - valued , site - specific profile , using the interactivity parameters derived by bastolla et al.(2004b ) .
the interactivity scale expresses the strength of effective interactions for each amino acid type , and its main component is the hydropathy .
the interactivity parameters correlate more strongly than other hydropathy scales with thermodynamic quantities like the thermodynamic effect of a mutation , with genomic quantities such as the genome size and the genome - wide gc content in bacterial genomes , and with the main component of substitution matrices , as we discuss in the present paper . by using the interactivity parameters and the pe as a structural indicator
, we recently derived analytic expressions to predict site - specific amino acid distributions ( porto et al . , 2004b ) .
amino acid distributions with a similar functional form have also been used previously by other authors ( koshi and goldstein , 1998 ; koshi et al .
1999 ; dokholyan et al . 2001 ; 2002 ) .
our predictions are in very good agreement with the distributions observed in the protein data bank .
here we consider simple mutational models at the nucleotide level , and we show that they improve the fit between predicted and observed distributions .
interestingly , the mutational model that best fits the observed data is one for which deleterious mutations eliminated by negative selection appear with almost the lowest rate .
we represent a protein structure as a @xmath0 binary contact matrix @xmath1 , where @xmath2 is the number of residues . the element @xmath1 is equal to one
if amino acids @xmath3 and @xmath4 are in contact in the native structure ( i.e. at least one pair of their respective heavy atoms is closer than @xmath5 ) , and zero otherwise .
the contact matrix is symmetric , so it has @xmath2 real eigenvalues .
the pe is the eigenvector corresponding to the maximum eigenvalue and is indicated by @xmath6 .
this vector maximizes the quadratic form @xmath7 with the constraint @xmath8 . in this sense
, @xmath9 can be interpreted as the effective connectivity of position @xmath3 , since positions with large @xmath9 are in contact with as many as possible positions @xmath4 with large @xmath10 .
all its components have the same sign , which we choose by convention to be positive .
moreover , if the contact matrix represents a single connected graph ( as for single - domain globular proteins ) , the information contained in the principal eigenvector is sufficient to reconstruct the whole contact matrix ( porto et al .
, 2004a ) .
as we will review in this paper , the pe provides a convenient vectorial representation of a protein structure to be used in evolutionary studies .
the effective free energy for a sequence @xmath11 in configuration @xmath12 is estimated through an effective contact free energy function @xmath13 , @xmath14 where @xmath15 is a @xmath16 symmetric matrix whose element @xmath17 represents the effective interaction , in units of @xmath18 , of amino acids @xmath19 and @xmath20 ; we use the interaction matrix derived by bastolla et al.(2001 ) . for this interaction matrix the free energy function , eq . ( [ eq : energy ] ) , is lower for the native structure than for decoys generated by threading ( bastolla et al . , 2001 ) . the effective energy function , eq .
( [ eq : energy ] ) , can be approximated through the main component of its spectral decomposition @xmath21 , @xmath22 where @xmath23 is the largest eigenvalue ( in absolute value ) of the matrix @xmath17 and @xmath24 is the corresponding eigenvector .
the latter is mainly due to the contribution of hydrophobic interactions , as it is well known ( casari et al . , 1992 ; li et al . , 1997
we therefore call the @xmath2-dimensional vector @xmath25 the hydrophobicity profile ( hp ) of sequence @xmath11 ( bastolla et al . , 2004b ) .
the 20 parameters @xmath24 obtained from the principal eigenvector of the interaction matrix are called _ interactivity _ parameters ( ih ) .
we will also use an optimized interactivity scale that was obtained by bastolla et al .
( 2004b ) by optimizing the correlation between the hp and the pe over a large set of proteins . since the pe represents an effective connectivity , we indicate the latter scale with the abbreviation ch .
we consider eleven hydropathy scales , including the two interactivity scales described above .
they are : ( 1 ) the kd hydropathy scale , derived to identify trans - membrane helices using diverse experimental data ( kyte and doolittle , 1982 ) ; ( 2 ) the l76 hydropathy scale , which was derived by using experimental data and theoretical calculations ( levitt , 1976 ) ; ( 3 ) the r88 hydropathy scale , which is based on the transfer of solutes from water to alkane solvents ( roseman , 1988 ) ; ( 4 ) the augmented whilmey - white ( ww ) hydropathy scale , derived to improve recognition of trans - membrane helices ( jayasinghe et al . , 2001 ) ; ( 5 ) the g98 classification of amino acids into polar , hydrophobic , and amphiphylic classes , adopted by gu et al . ( 1998 ) to investigate the relationship between the hydrophobicity of a protein and the nucleotide composition of the corresponding gene ; ( 6 ) the mp hydropathy scale , derived from statistical properties of globular proteins ( manavalan and ponnuswamy , 1978 ) ; ( 7 ) the av hydropathy scale derived by averaging 127 normalized hydropathy scales published in the literature ( palliser and parry , 2001 ) ; ( 8) the fp hydropathy scale , derived from the experimental measurement of octanol / water partition coefficients ( fauchere and pliska , 1983 ) ; ( 9 ) the z04 scale , also called buriability , proposed by zhou and zhou ( 2004 ) ; ( 10 ) the interaction scale ih , obtained from the main eigenvector of the interaction matrix @xmath17 used in this work ( bastolla et al .
, 2004b ) ; ( 11 ) the optimized interactivity scale , or connectivity scale ch , which maximizes the correlation with the principal eigenvectors of protein contact matrices for a non - redundant set of protein data bank ( pdb ) structures ( bastolla et al . , 2004b ) .
all these scales have pairwise correlation coefficients ranging from a minimum of @xmath26 ( between the kd and l76 scales ) to a maximum of @xmath27 ( between ih and ch scales ) . in the structurally constrained neutral ( scn ) model ( bastolla et al . , 1999 , 2002 , 2003a , 2003b ) ,
starting from the protein sequence in the pdb , amino acid mutations are randomly proposed and accepted according to a stability criterion based on an effective model of protein folding ( bastolla et al . , 2000 , 2001 ) .
this `` structural '' approach has been pioneered by schuster and co - workers , with a series of studies of neutral networks of rna secondary structures ( schuster et al .
, 1994 ; huynen et al . , 1996 ; fontana and schuster , 1998 ) and
it has been applied to proteins by several groups ( gutin et al . , 1995 ;
bornberg - bauer , 1997 ; bornberg - bauer and chan , 1999 ; babajide et al . , 1997 ; govindarajan and goldstein , 1998 ; bussemaker et al . , 1997 ;
tiana et al . , 1998 ; mirny and shakhnovich , 1998 ; dokholyan and shakhnovich , 2001 ; parisi and echave , 2001 ) .
for testing the stability of a protein conformation , we use two computational parameters : ( i ) the effective energy per residue , @xmath28 , eq .
( [ eq : energy ] ) , where @xmath2 is the protein length .
this quantity correlates with the folding free energy per residue for a set of 18 small proteins that are folding with two - states thermodynamics ( correlation coefficient @xmath29 ; u. bastolla , unpublished result ) ; ( ii ) the normalized energy gap @xmath30 , which characterizes fast folding model sequences ( bastolla et al . , 1998 ) with well correlated energy landscapes ( bryngelson and wolynes , 1987 ; goldstein et al . , 1992
; abkevich et al . , 1994 ; gutin et al . , 1995 ; klimov and thirumalai , 1996 ) . in the scn model , a mutated sequence
is considered thermodynamically stable if both computational parameters are above predetermined thresholds ( bastolla et al . , 2003a ) .
simulations of the scn model were performed for several protein folds .
they all show the same two features : ( 1 ) after a very long evolutionary time , the protein sequence looses all recognizable similarity with the starting sequence , despite conserving the stability of its native fold .
( 2 ) structural conservation varies across protein positions .
the main structural determinant of evolutionary conservation is the principal eigenvector of the contact matrix .
we obtain site - specific amino acid distributions from the pdb as follows .
first , we select a non - redundant set of single - domain globular proteins , i.e. sequences with more than 90% homology with other chains in the pdb are excluded . a non - redundant list of proteins is given in the file ` nrdb90 ` , available at the pdb web site .
qualitatively equivalent results were obtained using the list selected with the method by hobohm and sanders ( 1994 ) with the much lower threshold of 25% sequence identity .
globular proteins are selected through the condition that the fraction of contacts per residue should be larger than a length - dependent threshold , @xmath31 .
this functional form represents the scaling of the number of contacts in globular proteins as a function of chain length ( the factor @xmath32 comes from the surface to volume ratio ) , and the two parameters are chosen so as to eliminate outliers with respect to the general trend , which are mainly non - globular structures .
single - domain proteins are selected by imposing the normalized variance of the pe components to be smaller than a threshold , @xmath33 .
multi - domain proteins have large pe components inside their main domains and small components outside them .
the pe components would be exactly zero outside the main domain if the domains do not form contacts .
therefore , multi - domain proteins are characterized by a larger normalized variance of pe components with respect to single - domain ones .
the threshold of @xmath34 eliminates most of the known multi - domain proteins and very few of the known single - domain proteins ( data not shown ) .
the proteins are divided into groups according to their length .
a first group comprises proteins with less than @xmath35 amino acids and contains about @xmath36 sites .
a second group comprises proteins of length between @xmath37 and @xmath38 amino acids and contains @xmath39 sites .
these two groups are joined to improve the statistics in cases where we are not interested in length dependence .
a third group comprises proteins with length between @xmath40 and @xmath41 amino acids and contains @xmath42 sites . within each group ,
sites are divided in structural classes depending on the value of their normalized pe component , @xmath43 , using a bin width of @xmath44 . for each structural class
, the distribution of the @xmath45 types of amino acids is obtained .
we refer to this distribution with the symbol @xmath46 , where @xmath19 indicates one of the 20 types of amino acids .
the jensen - shannon divergence of two given distributions @xmath47 and @xmath48 is defined as @xmath49 , \end{split}\ ] ] where @xmath50/2 $ ] .
this measure is equivalent to the difference of the entropy of the ` average distribution ' @xmath51 minus the average entropy of distributions @xmath52 and @xmath53 . it is positive , symmetric , and takes values between zero and one .
this information - theoretical distance measure , however , does not take into account the physico - chemical similarities between amino acids . in the present context , it is natural to focus our attention on the amino acid interactivity @xmath24 , which determines the predicted site - specific distributions .
predicted distributions are defined imposing the value of their first moment , so it is natural to define the @xmath54 distance as the squared difference between the second moments @xmath55 \right\}^2 \\ & \equiv \left ( [ h^2]_p - [ h^2]_q \right)^2 .
\end{split}\ ] ] we average the distance measures between observed and predicted distributions for different values of the structural parameter @xmath43 . in the case of the @xmath54 distance ,
we normalize this measure in order to reduce its dependence on the scale of the hydrophobicity parameters .
we choose the normalization @xmath56_p - [ h^2_i]_q\right)^2 \right>_i } { \left < [ h^2_i]_p \right>_i \ , \left < [ h^2_i]_q \right>_i } , \ ] ] where angular brackets define the average over positions @xmath3 .
the procedure described by henikoff and henikoff ( 1992 ) for calculating substitution matrices starts by generating a multiple sequence alignment and counting the amino acid distribution at each position , @xmath57 . for large number of aligned positions ,
the logarithm of odds matrix that they define can be approximated as @xmath58 , \ ] ] where the angular brackets indicate an average over aligned positions @xmath3 .
using eq .
( [ eq : substmat ] ) , we calculate the substitution matrix generated by aligning positions of proteins in the pdb having the same value of the normalized pe , @xmath43 , in an interval of width @xmath44 .
usually , the eigenvectors @xmath59 of a @xmath0 symmetric matrix @xmath60 are normalized such that their squared components sum to one , @xmath61 .
in this case , the spectral decomposition of the matrix can be expressed as @xmath62 the relative importance of each eigenvector , however , does not only depend on the corresponding eigenvalue @xmath63 , but also on the average eigenvector component . to make this dependence more explicit , it is convenient to transform the eigenvectors in such a way that their average component is zero and their variance is one , _
i.e. _ by defining new vectors @xmath64 .
the spectral decomposition now reads @xmath65 , \end{split}\ ] ] where @xmath66 .
these satisfy the normalization @xmath67 .
the first term in the square brackets is the same for all pairs @xmath3 and @xmath4 .
the second term is proportional to the sum of the normalized eigenvector components , and the third term is proportional to their product .
therefore , we define two combinations of the eigenvalues and the mean eigenvector components as @xmath68 and @xmath69
it is well known that hydrophobicity is one of the main determinants of protein thermodynamics .
gromiha et al .
( 1999 ) showed that properties reflecting hydrophobicity correlate with the stability of buried mutations .
we tested the ability of various hydrophobicity scales to predict the average effect of single amino acid mutations ( bastolla , unpublished ) .
we use a database of more than @xmath70 mutants , selected by guerois et al.(2002 ) .
we average the change in folding free energy for each amino acid substitution present in this database with more than four examples .
these average changes are correlated with the difference of hydrophobicity parameters for various hydropathy scales .
the rationale for testing these correlations is that , from eq .
( [ eq : hydro ] ) , it follows that the difference between the effective unfolding free energy of a protein mutated at position @xmath3 and the wild - type is proportional to @xmath71 \ , \sum_j c_{ij } \ , h(a_j)$ ] . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ]
lastly , we solve eq .
( [ eq : trans ] ) numerically for a mutation matrix obtained by considering all nucleotide mutations as equiprobable .
this process has a stationary distribution that is almost proportional to the number of codons of each amino acid .
accordingly , its prediction coincides almost perfectly with the one obtained by choosing @xmath72 proportional to the number of codons , even if detailed balance is not assumed .
this result justifies the approximation of considering matrices that satisfy detailed balance .
results corresponding to various models are summarized in table [ tab : w ] . using the mutation matrix @xmath73 and the fixation probability eq .
( [ eq - fix ] ) , we can compute the equilibrium fraction of mutations which are eliminated by negative selection at site @xmath3 with boltzmann parameter @xmath74 ( we omit the index @xmath3 for simplicity ) , @xmath75 the computation of this quantity requires to know the mutation matrix , whereas our assumption of detailed balance only requires to specify the stationary nucleotides frequencies under no selection ( @xmath76 ) .
therefore , we evalutate the rejection rate through the average of the absolute value of the boltzmann parameters @xmath77 . among the models that we considered , @xmath77 is lowest for the more elaborate model taking into account nucleotide frequencies . among all possible nucleotide frequencies ,
the lowest value of @xmath77 is @xmath78 , and it is attained at nucleotide frequencies close to those giving the optimal fit , @xmath79 , @xmath80 , @xmath81 , and @xmath82 .
even lower values of @xmath83 and better fits can be obtained deriving the @xmath72 from the amino acid frequencies observed at specific values of the pe component ( the minimum value of latexmath:[$\left < @xmath72 coincide with the frequencies observed at @xmath85 ) , but these values are ad - hoc since they are not the result of a mutational model .
these results suggest that mutation frequencies are set to almost optimize the mutational load .
kinjo and nishikawa ( 2004 ) recently calculated substitution matrices derived from protein alignments with various values of the sequence identity of the aligned proteins .
qualitatively , these substitution matrices measure the tendency of two residue of types @xmath19 and @xmath20 to co - occur at aligned sites .
kinjo and nishikawa noticed that the eigenvector corresponding to the largest positive eigenvalue of the substitution matrices is correlated with the hydrophobicity , and that the corresponding eigenvalue increases with decreasing sequence alignment , corresponding to longer evolutionary time .
we examine the substitution matrices calculated by kinjo and nishikawa for various sequence identities , and we consider for comparison the matrix blosum62 ( henikoff and henikoff , 1992 ) . we also obtain a new substitution matrix from the alignment of pdb structures based on the normalized pe components , as described in materials and methods .
this latter substitution matrix can also be predicted analytically using the formulas for site - specific distributions given above .
its non - diagonal elements are correlated with the corresponding elements of the blosum62 matrix , with @xmath86 .
however its diagonal elements are much smaller than the corresponding blosum elements , since all 20 amino acids can be found with fairly large frequencies at each aligned position .
we calculate the correlation of the eigenvector corresponding to the largest positive eigenvalue of these substitution matrices with the hydrophobicity scales described in materials and methods . for almost all matrices , the largest correlation ( with @xmath87 to @xmath88 )
is found for the optimized interactivity parameters , but the correlation is very high also for other scales , see table [ tab : hydro ] .
we then computed the normalized eigenvalue of this eigenvector , eq .
( [ eq : ell - prod ] ) . from its behavior as a function of sequence identity
( see fig .
[ fig : subst_mat ] ) one sees that the contribution of the hydrophobicity to the substitution matrices becomes more important for lower identities of the aligned proteins , _
i.e. _ for increased evolutionary time .
the blosum matrix corresponds to an intermediate value , whereas the matrix obtained from aligned pdb structure is completely dominated by hydrophobicity , as expected from eq .
( [ eq : substmat ] ) .
this matrix corresponds to the limit of very long evolutionary time , where all sequence similarity is lost , apart for that originated by the common structural environment expressed by the pe .
therefore , this study of substitution matrices further supports our results on the relationship between the pe and sequence evolution .
it is well known that the hydrophobicity is a major determinant of protein stability and evolution .
we have shown that , among several hydropathy scales , the optimized interactivity scale ( bastolla et al . , 2004b ) is the one best correlated with various thermodynamic and genomic quantities .
the influence of the hydrophobicity on protein stability has two sides : on the one hand , the more hydrophobic a protein is , the more stable is with respect to unfolding ; on the other hand it is less stable with respect to misfolding , so that a careful tuning of the hydrophobic content is needed ( bastolla et al . , 2004a ) .
close to the optimal hydrophobicity , small variations have complex effects that might reduce their impact on fitness .
this is probably the reason why the mutational bias is able to influence hydrophobicity and hence protein folding thermodynamics .
large variations of the mean hydrophobicity , however , are not tolerated in the course of evolution . in order to achieve folding stability ,
we have predicted that the hydrophobicity should be distributed within the protein sequence in a site - specific manner , which is mainly influenced by a site - specific indicator
the principal eigenvector of the contact matrix ( pe ) . from the hypothesis that the correlation between the hydrophobicity and the pe is the only condition imposed by folding stability on sequence evolution , we predicted analytically site - specific amino acid distributions to be of boltzmann type , in agreement with simulations of the scn model of protein evolution and with structural alignments of sites with equivalent pe taken from the pdb .
finally , we have discussed how simple mutational models improve the fit between predicted and observed site - specific distributions .
interestingly , the model that provides the best fit is also one with almost the lowest rate of rejected mutations .
this observation is consistent with the suggestions that error correction systems might have evolved to minimize the impact of mutations on protein stability .
abkevich , v.i . ,
gutin , a.m. , shakhnovich , e.i .
free energy landscapes for protein folding kinetics - intermediates , traps and multiple pathways in theory and lattice model simulations .
j. chem .
phys . * 101*:6052 - 6062 .
casari g. , sippl , m.j .
structure - derived hydrophobic potential .
hydrophobic potential derived from x - ray structures of globular proteins is able to identify native folds .
j. mol .
biol . * 224*:725 - 732 .
gromiha , m.m . ,
oobatake , m. , kono , h. , uedaira , h. , sarai , a. 1999 .
role of structural and sequence information in the prediction of protein stabililty changes : comparison between buried and partially buried mutations .
* 12*:549 - 555 .
porto , m. , roman , h.e . ,
vendruscolo , m. , bastolla , u. 2004b .
prediction of site - specific amino acid distributions and limits of divergent evolutionary changes in protein sequences .
, in print . | we review and further develop an analytical model that describes how thermodynamic constraints on the stability of the native state influence protein evolution in a site - specific manner . to this end
, we represent both protein sequences and protein structures as vectors : structures are represented by the principal eigenvector ( pe ) of the protein contact matrix , a quantity that resembles closely the effective connectivity of each site ; sequences are represented through the `` interactivity '' of each amino acid type , using novel parameters that are correlated with hydropathy scales .
these interactivity parameters are more strongly correlated than the other hydropathy scales that we examine with : ( 1 ) the change upon mutations of the unfolding free energy of proteins with two - states thermodynamics ; ( 2 ) genomic properties as the genome - size and the genome - wide gc content ; ( 3 ) the main eigenvectors of the substitution matrices .
the evolutionary average of the interactivity vector correlates very strongly with the pe of a protein structure . using this result ,
we derive an analytic expression for site - specific distributions of amino acids across protein families in the form of boltzmann distributions whose `` inverse temperature '' is a function of the pe component .
we show that our predictions are in agreement with site - specific amino acid distributions obtained from the protein data bank , and we determine the mutational model that best fits the observed site - specific amino acid distributions .
interestingly , the optimal model almost minimizes the rate at which deleterious mutations are eliminated by natural selection . |
there is a current interest in the thermodynamics and statistical mechanics of fluctuating systems in contact with heat reservoirs and driven by external forces .
the current focus stems from the recent possibility of direct manipulation of nano - systems and bio - molecules .
these techniques permit direct experimental access to the probability distribution functions for the work or for the heat exchanged with the environment @xcite .
these methods have also yielded access to the experimental verification of the recent fluctuation theorems which relate the probability of observing entropy - generated trajectories with that of observing entropy - consuming trajectories @xcite . in recent works we studied the motion of a brownian particle in a general potential with a view to the distribution function for the heat exchange with the surroundings
@xcite and a single bound brownian particle driven by two heat reservoirs @xcite . in the present paper we consider the harmonic chain driven by heat reservoirs at temperatures @xmath1 and @xmath2 @xcite . here the distribution of positions and momenta is given by a gaussian form with a correlation matrix with elements given by the static position and momentum correlations @xcite
. owing to the current interest in fluctuation theorems the linear chain has recently been addressed again by saito and dhar @xcite and by kundu et al .
@xcite ; see also @xcite for a treatment in the quantum regime . using a path integral formulation , fourier series , and analyzing the resulting energy transmission matrices , these authors derive an expression for the cumulant generating function , in the following denoted cgf , for the heat transfer in terms of a transmission green s function @xmath3 .
the large deviation function for the heat transfer , denoted ldf , then follows by a legendre transformation of the cgf ; for definitions , see later .
the expression is in accordance with the fluctuation theorem @xcite . in the present paper
we consider four issues : i ) the cgf for the harmonic chain , ii ) the ldf for the chain and the exponential tails in the heat distribution , iii ) the cgf for an extension of a model by derrida and brunet @xcite , iv ) the cgf for the harmonic chain in the large @xmath0 limit , where @xmath0 is the number of particles , and v ) a derivation of the fluctuation theorem on the basis of a fokker - planck description .
for the purpose of the analysis in ii ) - iv ) we have within a langevin scheme performed a calculation of the cgf , including explicit expressions for the transmission green function . at the technical level
we , moreover , unlike kundu et al .
@xcite , make use of fourier transforms throughout the calculation and diagonalize explicitly @xmath3 expressing the cgf in terms of the eigenvalues . for the benefit of the reader and the continuity of the paper we have chosen to include this analysis in the main part of the paper .
we discuss the tails in the heat distribution and exemplify this feature both for the extended derrida - brunet model and the @xmath0 particle chain .
we consider the cgf and ldf for an extension of a single particle model suggested by derrida and brunet @xcite and the cgf in the two - particle case .
we , moreover , analyze the asymptotic large @xmath0 limit and present a closed expression for the cgf .
finally , as a related and more formal issue we present a derivation of the fluctuation theorem on the basis of a fokker - planck description of a chain . as a bonus we are able to prove that the fluctuation theorem holds for chains with general interaction potentials and with several heath baths at different temperatures . for reference we present below the results of saito and dhar @xcite and kundu et al .
@xcite , also presented in the present paper . denoting the model - dependent transmission greens function by @xmath3 , the cgf @xmath4 for the characteristic function for the transferred heat @xmath5 in the time interval @xmath6 ,
is given by the following expressions : @xmath7 , \label{2 } \\ & & f(\lambda)=t_1t_n\lambda(1/t_1 - 1/t_n-\lambda ) .
\label{3}\end{aligned}\ ] ] here @xmath1 and @xmath2 denote the reservoir temperatures and the form of @xmath8 ensures the validity of the fluctuation theorem @xmath9 as a new result we present below the cgf in the asymptotic large @xmath0 limit . here
@xmath10 denotes the reservoir damping and @xmath11 the spring constant .
the cgf is given by @xmath12 .
\label{4}\end{aligned}\ ] ] finally , we note that the mathematical background for the present paper is provided by large deviation theory , see refs .
@xcite .
the paper is organized in the following manner . in sec .
[ chain ] we present the harmonic chain . in sec .
[ analysis ] we set up the necessary analysis . in sec .
[ large ] we present a derivation of the cgf .
the general properties of such a function are discussed in sec .
[ discuss ] , where we also consider the tails of the heat distribution , the specific cases of a bound brownian particle , a two - particle chain , and the large @xmath0 limit . in sec .
[ gen ] we discuss a generalization of the fluctuation theorem . in sec .
[ sum ] we present a summary and a conclusion .
the dynamics of a unit mass harmonic chain composed of @xmath0 particles and , moreover , attached to a wall or substrate , is governed by the hamiltonian @xmath13 where @xmath14 and @xmath15 denotes displacements and momenta , respectively ; @xmath11 is the spring constant .
the equation of motion for the bulk particles and the end particles driven by the heat reservoirs at temperatures @xmath1 and @xmath2 with associated damping @xmath10 are given by @xmath16 with noise correlations and strengths @xmath17 in equilibrium @xmath18 and detailed balance implies @xmath19 , where @xmath20 is the common temperature of the reservoirs
. focussing on the reservoir at temperature @xmath1 the fluctuating force is given by @xmath21 and , correspondingly , the rate of work or heat flux has the form , denoting @xmath22 , @xmath23 the central quantity in the analysis is , however , the total heat transmitted during a finite time interval @xmath6 , i.e. , @xmath24 note that strictly speaking only the time scaled heat @xmath25 has large deviation properties , see e.g. refs .
@xcite .
the heat @xmath5 is fluctuating and the issue is to determine its probability distribution @xmath26 ; here @xmath27 denotes an average with respect to @xmath28 and @xmath29 . in terms of the characteristic function
@xmath30 we have by a laplace transform @xcite @xmath31 the chain attached to the substrate at the ends and driven by heat reservoirs is depicted in fig .
the heat reservoirs drive the chain into a stationary state .
since the heat is transported ballistically the only damping mechanism is associated with the heat reservoirs and the only time scale is given by @xmath32 .
consequently , at long times compared with @xmath32 we can neglect the initial preparation of the chain and analyze the problems in terms of fourier transforms . thus introducing the fourier transform @xmath33 the equations of motion ( [ eq1 ] ) to ( [ eq4 ] ) and noise correlations ( [ n1 ] ) to ( [ n2 ] )
take the form @xmath34 here the inverse green s function @xmath35 is a symmetrical tridiagonal matrix with elements @xmath36 where @xmath37 note that for a free chain we have @xmath38 . propagating bulk solutions
have the form @xmath39 where @xmath40 is confined to the first brillouin zone @xmath41 and , correspondingly , @xmath42 . imposing the noisy drive we readily determine the coefficients @xmath43 and @xmath44 and
infer the solutions @xmath45 where the green function components are given by @xmath46 the displacement @xmath14 is thus driven by stochastically excited lattice waves ( phonons ) propagating towards the site from the end points ; @xmath47 yield the damped mode spectrum .
also , from the definition of @xmath48 we deduce the relationship @xmath49 and by multiplication the schwinger identity @xcite @xmath50.\label{sc}\end{aligned}\ ] ] in the absence of the heat reservoirs energy is conserved , i.e. , @xmath51 , where @xmath52 is given by ( [ ham ] ) .
coupling the reservoirs to the chain we have @xmath53 , where @xmath54 is the heat flux from the reservoir at temperature @xmath2 .
averaging we have for the mean heat fluxes @xmath55 , expressing the energy balance ; the mean input flux at @xmath56 is equal to the mean output flux at @xmath57 . using ( [ hf ] ) , inserting ( [ sol ] ) , averaging over the noises ( [ n11 ] ) and ( [ n22 ] ) , using the properties of the green s function ( [ g1 ] ) and ( [ gn ] ) , and the identity ( [ sc ] )
, we obtain for the mean transferred heat in time @xmath6 @xmath58 here the central model dependent quantity is the end - to - end greens function @xmath59 ; we note that the mean heat vanishes for @xmath60 .
we also note that the transferred mean heat rate @xmath61 is given by @xmath62 see ref .
@xcite . here
@xmath63 is the average kinetic temperature of the first particle in the steady state ; note that in equilibrium @xmath64 and @xmath65 in accordance with the equipartition theorem @xcite .
the relation ( [ qbar ] ) follows from ( [ mean ] ) by inserting the greens function solution of the equations of motion and using the identity ( [ sc ] ) .
the expression ( [ qbar ] ) also follows from the equivalent fokker - planck approach to the harmonic chain which we discuss below . considering the definition of the n - th moment of the heat transfer in time @xmath6 @xmath66 and referring to ( [ fp2 ] ) in sec .
[ gen ] , the fokker - planck equation for the joint distribution @xmath67 in the case of two reservoirs implies @xmath68 these equations of motion are part of a hierarchy relating the n - th moment to correlations of the lower moments with @xmath69 and have to be completed by equations of motions for the correlations @xmath70 . without further assumptions this hierarchy will in general not terminate and simply represents a reformulation .
we note , however , that for the first moment for @xmath56 the second term in ( [ momeq ] ) vanishes and we obtain a closed equation yielding ( [ qbar ] ) . for the fluctuating heat transferred in time @xmath6 we obtain , using ( [ h ] ) and inserting ( [ sol ] )
, the expression @xmath71 the heat transfer is a fluctuating quantity depending bilinearly on the reservoir noises @xmath28 and @xmath29 .
the matrix elements in the symmetrical form ( [ heat ] ) are given by @xmath72 where we have introduced the notation @xmath73 we note that ( [ sc ] ) implies @xmath74 .
\label{sc2}\end{aligned}\ ] ] the dependence on the transfer time @xmath6 is embodied in the function @xmath75 for later purposes we also note that @xmath76 at this stage our calculation differs from kundu et al .
@xcite in that we use a fourier transform instead of a fourier series in the expression ( [ heat ] ) for the fluctuating heat .
the dependence on the transfer time @xmath6 is then incorporated in the function @xmath77 .
for large @xmath6 the mean heat @xmath78 given by ( [ mean ] ) grows linearly with time . analyzing the higher cumulants @xmath79 , i.e. , @xmath80 , etc .
, by averaging over the noise and applying wick s theorem @xcite , it also follows that they likewise increase linearly with time , i.e. , @xmath81 for large @xmath6 .
we thus infer from the cumulant expansion of the characteristic function @xcite , @xmath82 that for large @xmath6 @xmath83 where @xmath4 is the cumulant generating function , denoted cgf .
the cgf characterizes the long time heat distribution . from the cumulant expansion ( [ cum ] )
we obtain the relationship @xmath84 here the definition ( [ char ] ) for @xmath85 implies @xmath86 in case the fluctuation theorem is valid we , moreover , have the symmetry @xmath87 turning to the evaluation of @xmath4 we average @xmath88 with respect to the noises @xmath28 and @xmath29 . in matrix form the gaussian noise distribution has the form @xmath89 where @xmath90 and the inverse noise matrix is given by @xmath91 noting from ( [ heat ] ) that @xmath5 is bilinear in @xmath92 and using the identities @xcite @xmath93 we obtain for the cgf @xmath94 in the remaining part of this section the present calculation differs from kundu et al .
@xcite in that we owing to the nondiagonal character of @xmath95 must expand @xmath96 in order to implement the large @xmath6 limit . thus expanding the log according to @xmath97 and tracing term by term we have @xmath98 and
the issue is to determine @xmath99 and complete the sum . from ( [ heat ] ) we obtain @xmath100 where @xmath101 . inserting ( [ f ] )
we notice that since @xmath102 the exponential factors in the product of f functions combine yielding a unit factor .
we thus only have to retain the sine part , i.e. , @xmath103 .
using ( [ f1 ] ) and ( [ f2 ] ) we have for @xmath104 @xmath105 for large @xmath6 the function @xmath77 oscillates rapidly as a function of @xmath106 and we have approximately @xmath107 , i.e. , the effective integration range in @xmath106 space is confined to the domain @xmath108 and only one @xmath106 integration remains . using @xmath109 inspection readily yields @xmath110 and the cgf takes the form @xmath111 in order to complete the calculation we diagonalize the two - by - two matrix @xmath112 . denoting the eigenvalues by @xmath113 and @xmath113 we have @xmath114 and reconstructing the log we obtain for @xmath4 @xmath115 , \label{ldf33}\end{aligned}\ ] ] the eigenvalues @xmath113 and @xmath113 are determined by the condition @xmath116 , i.e. , @xmath117 yielding the quadratic equation @xmath118 with roots @xmath119 and @xmath120 . in particular @xmath121 using the identity ( [ sc2 ] ) we obtain the reduced expressions @xmath122 i.e. , @xmath123 and for the cgf @xmath124 . \label{ldf6}\end{aligned}\ ] ] finally , inserting ( [ s1 ] ) and ( [ s2 ] ) the cgf can be expressed in the form @xmath125 , \label{ldff}\end{aligned}\ ] ] where @xmath126 this expression is in agreement with kundu et al .
@xcite . here
the form of @xmath8 ensures that the fluctuation theorem ( [ ft ] ) holds .
the deterministic dynamics of the chain is entirely embodied in the momentum green s function @xmath127 .
here we discuss four issues : i ) the branch cut structure in @xmath4 and ensuing exponential tails in the heat distribution @xmath128 , ii ) a single bound brownian particle coupled to two reservoirs , iii ) a two particle chain coupled to heat reservoirs , and iv ) an asymptotic expression for @xmath4 in the large @xmath0 limit . by inspection of the general expression ( [ ldff ] ) for the cgf
we infer that @xmath4 has the form of a downward convex function passing through the origin @xmath129 due to normalization and through @xmath130 owing to the fluctuation theorem .
since the argument in the log in ( [ ldff ] ) must be positive we infer the condition @xmath131 where @xmath132 is the maximum value of @xmath133 in the @xmath106 range . by means of algebraic and trigonometric manipulations it can be shown that @xmath134 is bounded by @xmath135 , for details see appendix [ app1 ] , and consequently , @xmath136 . by analyzing the expression for @xmath8 in ( [ ff ] )
one easily finds that this bound is satisfied for @xmath137 , where the branch points @xmath138 in @xmath4 are given by @xmath139 in fig .
[ fig2 ] we have depicted the cgf given by ( [ ldff ] ) for the case @xmath140 , @xmath141 , @xmath142 , @xmath143 , and @xmath144 . at large times
the heat distribution function follows from ( [ dh ] ) , i.e. , @xmath145 and the rate function or large deviation function @xmath146 is given by @xmath147 since @xmath4 is differentiable , strictly convex , and steep at the boundaries the grtner - ellis theorem @xcite implies that the ldf is given by the legendre transform @xmath148 or @xmath149 where @xmath150 is determined by @xmath151 and we find the ldf @xmath152 for @xmath146 the fluctuation theorem has the form @xmath153 note that the ldf also follows from a heuristic saddle point argument , see @xcite . in fig .
[ fig3 ] we have depicted @xmath154 for the case @xmath140 , @xmath141 , @xmath142 , @xmath143 , and @xmath144 . replacing @xmath4 by the parabolic approximation @xmath155 where @xmath156 is given by ( [ qbar ] ) we obtain for @xmath146 @xmath157 in accordance with ( [ ft2 ] ) . for the heat distribution
we obtain the displaced gaussian @xmath158 ; this also follows from general large deviation theory @xcite .
deforming the contour in the integral ( [ dh2 ] ) to pass along the real axis we pick up branch cut contributions in @xmath4 .
heuristically , we conclude that for large @xmath159 @xmath160 where @xmath161 and @xmath162 have been defined above .
this also follows directly form the legendre transformation since @xmath4 is defined on a compact support .
the linear behavior is confirmed by the plot of @xmath146 for our particular choice of the parameter set , see fig .
the heat distribution thus exhibits exponential tails for large @xmath159 , i.e. , @xmath163 with @xmath161 and @xmath162 given by ( [ brp1 ] ) and ( [ brp2 ] ) .
it is interesting that the tails are determined only by the reservoir temperatures .
finally , we note that the exponential tails in @xmath164 also follows from large deviation theory since @xmath96 is bounded by @xmath138 , see refs .
@xcite . in an interesting paper
derrida and brunet @xcite considered a single brownian particle driven by two reservoirs at distinct temperatures and presented an explicit expression for the cgf @xmath4 .
this toy model has also been discussed by visco @xcite who considered next leading term and the role of initial conditions ; see also farago @xcite . in a previous paper @xcite we considered an extension of this model to the case of a single particle attached harmonically to a substrate with spring constant @xmath11 using the simple method devised by derrida and brunet .
we found that the cgf is _ independent _ of @xmath11 , indicating that the deterministic character of the spring does not influence the statistical properties of the long time heat transfer .
here we consider as an illustration the same problem within the present scheme and recover a cgf independent of @xmath11 .
the configuration is shown in fig .
[ fig4 ] . associating the damping constants @xmath165 and @xmath166 with the two reservoirs the equation of motion take the form @xmath167 with noise correlations @xmath168 in fourier space we obtain the solution @xmath169 where @xmath170 and @xmath171 note that @xmath127 satisfies the schwinger identity ( [ sc2 ] ) , i.e.
, @xmath172 the heat flux from the reservoir at temperature @xmath1 is @xmath173 and we obtain from ( [ solb ] ) , ( [ heat ] ) , and ( [ sc3 ] ) the diagonal matrix elements @xmath174 following the prescription in sec .
[ large ] the eigenvalue equation imply @xmath175 and we obtain the cgf @xmath176 , \label{ldfb}\end{aligned}\ ] ] where @xmath134 is given by ( [ ab2 ] ) and @xmath8 by ( [ ff ] ) . a straightforward evaluation of ( [ ldfb ] ) using the integral @xcite @xmath177 yields @xmath178 where @xmath179 further inspection shows , however , that the combination @xmath180 is _ independent _ of the spring constant @xmath11 , as already shown in @xcite , and we obtain @xmath181 . \label{ldfb3}\end{aligned}\ ] ] introducing @xmath8 , as defined in ( [ ff ] )
, we can express ( [ ldfb3 ] ) in the form @xmath182 where the branch points are given by @xmath183 .
\label{lamb}\end{aligned}\ ] ] we note that @xmath134 given by ( [ ab2 ] ) has a two - peak structure with maximum value @xmath184 at @xmath185 and that the expressions ( [ ldfb ] ) and ( [ lamb ] ) are in accordance with the general properties of @xmath4 discussed above . finally , using ( [ spc ] ) and ( [ pldf ] ) we obtain for the large deviation function @xmath146 , @xmath186 , \label{pldf2}\end{aligned}\ ] ] which yields a heat distribution @xmath164 in accordance with the general discussion in sec .
[ discuss ] a with exponential tails in the heat distribution ; for more details regarding this model , see @xcite .
as an illustration of the general scheme presented here we briefly consider the case of a chain composed of two particles ; the configuration is depicted in fig .
[ fig5 ] . setting @xmath187 we have from ( [ ff1 ] ) , ( [ gn ] ) , and ( [ a2 ] ) @xmath188 and
we obtain from the general expression ( [ ldff ] ) the cgf @xmath189 . \label{2ldfb}\end{aligned}\ ] ] we have been unable to reduce the expression ( [ 2ldfb ] ) further but note that for @xmath190 the ldf @xmath191 for all @xmath192 , corresponding to two independent equilibrium systems at temperatures @xmath1 and @xmath193 .
we also remark that decoupling the chain from the walls , corresponding to setting @xmath194 , we obtain @xmath195 . in the limit of a stiff chain ,
corresponding to @xmath196 , we have @xmath197 , i.e. , the case of a single unbound particle coupled to two reservoirs , see sec .
[ discuss ] b. in the limit of large @xmath0 we present below an asymptotic expression for the cgf . for general @xmath0
we have from ( [ dis ] ) , ( [ gn ] ) , ( [ d ] ) , and ( [ a22 ] ) @xmath198 which expanding the denominator can be expressed in the form @xmath199 where @xmath200 by inspection we note that @xmath134 displays an oscillatory structure with approximate period @xmath201 , reflecting the resonance structure of the propagating lattice waves in the chain .
the oscillations are modulated by the slowly varying functions of @xmath40 , @xmath202 and @xmath203 .
further inspection of ( [ bn2 ] ) shows that the maxima are given by @xmath204 where a little analysis implies that @xmath205 locks onto @xmath206 , corroborating the demonstration of the upper bound in appendix [ app1 ] .
the lower bound of the oscillatory structure is , correspondingly , given by the envelope @xmath207 the structure is for @xmath144 depicted in fig .
the positions of the maxima and minima are given by the implicit conditions @xmath208 and @xmath209 , respectively .
in @xmath40 space , using @xmath210 , and inserting ( [ bn2 ] ) , the cgf given by ( [ ldff ] ) takes the form @xmath211 . \label{2ldfb2}\end{aligned}\ ] ] in the large @xmath0 limit the rapid oscillations in @xmath212 allows us to integrate separately over each period . using the integral @xcite ,
see also ref .
@xcite , @xmath213 we thus obtain the following approximate form of @xmath212 , @xmath214 further , inserting @xmath215 and @xmath216 from ( [ l ] ) and ( [ k ] ) we obtain @xmath217 and for the cgf in the limit @xmath218 @xmath12 . \label{2ldfb3}\end{aligned}\ ] ] in fig .
[ fig6 ] we have for @xmath144 , @xmath142 , and @xmath143 depicted @xmath212 , @xmath219 , @xmath220 , and @xmath221 .
we note that @xmath221 smoothly interpolates over the oscillations in @xmath212 . in fig .
[ fig7 ] we depict @xmath4 as a function of @xmath96 for @xmath187 and for @xmath144 .
the other parameters are @xmath142 , @xmath143 , @xmath222 , and @xmath223 .
we note the excellent fit already for @xmath144 and the good approximation at small @xmath192 for @xmath187 .
the expression ( [ 2ldfb3 ] ) for @xmath4 is manifestly independent of @xmath0 in the large @xmath0 limit .
this implies according to ( [ cum2 ] ) that the cumulants and in particular the mean current also are independent of @xmath0 .
this signals that fourier s law is not valid for the harmonic chain , see e.g. ref .
we also note that the large @xmath0 limit does not correspond to the continuum limit ; we just increase the number of particles in the chain keeping the lattice distance fixed .
in secs . [ large ] and [ discuss ] we demonstrated the validity of the fluctuation theorem by an explicit evaluation of the cgf for the harmonic chain driven at the end points by heat reservoirs at distinct temperatures and considered , moreover , some special cases .
here we put these results in a more general framework by considering the fokker - planck equation for the characteristic function @xmath224 for long times @xmath225 and we obtain the differential equation @xmath226 expressing the fokker - planck equation for @xmath227 in the form @xmath228 we identify the cgf @xmath4 as the maximal eigenvalue of the fokker - planck operator @xmath215 .
the issue is thus to establish the fluctuation theorem symmetry for the maximal eigenvalue .
we aim at a generalization of the fluctuation theorem to the case of many heat reservoirs , see also @xcite . for that purpose we consider a setup where each particle in the chain couples to its own heat reservoir at temperature @xmath229 .
the configuration is depicted in fig .
[ fig8 ] . generalizing ( [ hf ] ) the heat flux to the n - th particle
is given by @xmath230 where the noise is correlated according to @xmath231 since the transfer of heat induces a change in the state of the system we must at the outset consider the joint distribution @xmath232 .
the heat distribution is then given by @xmath233 .
as discussed in ref .
@xcite , the fokker - planck equation for the joint distribution @xmath67 is derived by considering the heat @xmath234 as an independent dynamical variable , whose time evolution is governed by ( [ hfn ] ) .
noting that the noise appearing in this equation is correlated to the noise appearing in the equation of motion for the momenta ( [ eq1]-[eq4 ] ) one can write @xmath235 where the poisson bracket is given by @xmath236 ; \label{pb}\end{aligned}\ ] ] see also ref .
@xcite .
the structure of ( [ fp4 ] ) also allows a simple derivation of a generalized fluctuation theorem ; see also ref .
the first step is to perform a rotation " @xmath250 with respect to the m - th reservoir in combination with a time reversal operator @xmath251 and define the transformed fokker - planck operator @xmath252 in the next step we compare the operator @xmath253 with the adjoint operator @xmath254 . using @xmath255 and @xmath256 and shifting the laplace variables
@xmath257 it turns out that @xmath253 and @xmath254 become identical and we have the relationship @xmath258 where @xmath259 since @xmath241 is related to @xmath260 by a unitary transformation we infer that @xmath241 and @xmath261 have identical spectra and in particular identical maximal eigenvalues , i.e. , the same large deviation function , @xmath262 the expression ( [ ftg ] ) together with ( [ lam ] ) represents a generalization of the usual fluctuation theorem to many reservoirs . in the case of two reservoirs , setting @xmath263 , @xmath264 , and @xmath265 we obtain the usual fluctuation theorem ( [ ft ] ) .
we note that the above derivation holds for any time reversal invariant hamiltonian , i.e. , for any kind of interaction between the particles . in our derivation
we have also assumed that the maximal eigenvalue is positive .
thus , our proof of the fluctuation theorem is more general than the one given in ref .
@xcite , which is restricted to the harmonic chain only .
furthermore , our approach does not require a direct evaluation of the cgf , but is based only on the property of the dynamics , as expressed by the evolution operator @xmath241 .
the previous proof can be readily extended to the 3-d case , as long as @xmath241 has a form as in ( [ fp4 ] ) .
in this paper we have discussed a variety of issues regarding the noise driven harmonic chain . in secs . [ analysis ] and
[ large ] we performed a calculation of the cgf , recovering the results of kundu et al .
@xcite , but adding some more details for the purpose of our analysis . in sec
[ discuss ] we discussed the exponential tails in the heat distribution , the bound single particle model , and the two - particle chain case .
it is an interesting feature of the tails that the fall - off rate only depends on the noise features , i.e. , the reservoir temperatures , and not on the dynamical properties of the chain such as the spring constant @xmath11 . in the large @xmath0 limit
we have found an analytical form for the cgf which excellently interpolates the exact result .
this result is independent of @xmath0 signalling that fourierr law does not hold . finally , incorporating some of our results we have in sec .
[ gen ] within a fokker - planck description presented a generalization of fluctuation theorem to several reservoirs which holds for any interaction potential .
the fluctuation theorem simply emerges from a symmetry hidden in the fokker- planck operator and is therefore not restricted to the linear chain but also holds for a 3d system .
we are grateful to c. mejia - monasterio and f. van wijland for interesting discussions .
we also thank a. mossa , a. svane , u. poulsen , and g. bruun for useful discussions .
hf gratefully acknowledges financial support by the danish natural science research council under grant no .
09 - 072352 .
ai gratefully acknowledges financial support by lundbeck fonden .
here we analyze the modulus squared of the function @xmath266 where @xmath267 and demonstrate that it is bounded from above by @xmath206 , i.e.,@xmath268 . breaking up @xmath269 in real and imaginary parts , @xmath270}&=&-4 \kappa \gamma^2 \sin^2(p/2 ) \sin(n-1)p+
\kappa^2 \sin(n+1)p , \label{red } \\
\im{\left[{d(\omega)}\right]}&=&-4 \kappa^{3/2}\gamma \sin(p/2 ) \sin n p , \label{imd}\end{aligned}\ ] ] inserting @xmath134 , expressing @xmath106 in terms of @xmath40 , using ( [ omegap ] ) , and substituting ( [ red])-([imd ] ) , we rephrase the condition @xmath271 as @xmath272 } \nonumber \\ & & + ( -4\kappa\gamma^2\sin(p/2)\sin(n-1)p+\kappa^2\sin(n+1)p)^2 \ge 0.\end{aligned}\ ] ] expressing the @xmath203 in terms of @xmath273 , and rearranging terms , @xmath274 becomes @xmath275 which is non negative thus demonstrating our assertion .
the values of @xmath40 for which @xmath276 correspond to the points of maximum for @xmath134 , with the exception of @xmath277 where @xmath47 .
so , @xmath134 has n-1 maxima where @xmath278 , see fig .
[ fig9 ] . and @xmath2 .
the chain is attached to walls or substrates at the ends .
the total heat transmitted to the @xmath56 and @xmath57 particles are denoted @xmath279 and @xmath280 , respectively .
the spring constant is denoted @xmath11 . ]
[ ct][ct][1.5]@xmath4 [ ct][ct][1.5]@xmath192 [ ct][ct][1.5]@xmath162 [ ct][ct][1.5]@xmath161 , as given by ( [ ldff ] ) for @xmath140 , @xmath141 , @xmath142 , @xmath143 , @xmath144 .
inset : zoom of the plot for small value of @xmath192.,title="fig:",width=566 ] [ ct][ct][1.5]@xmath154 [ ct][ct][1.5]@xmath281 as a function of @xmath281 , as given by ( [ pldf ] ) for @xmath140 , @xmath141 , @xmath142 , @xmath143 , @xmath144 .
dotted line : parabolic approximation , ( [ quad2 ] ) . dashed and dotted - dashed line : linear regime for @xmath282 , the slopes are @xmath283 and @xmath284 , respectively.,title="fig:",width=566 ] and @xmath193 . the chain is attached to walls or substrates at the ends .
the heat transferred to the particles is denoted @xmath279 and @xmath285 , respectively .
the particle is attached to a substrate with a harmonic spring with force constant @xmath11 . ]
[ ct][ct][1.5]@xmath286 [ ct][ct][1.5]@xmath40 given by ( [ bn2 ] ) as function of @xmath40 in the range @xmath287 for @xmath144 , @xmath142 , and @xmath143 ( blue ) .
we also show the maximum value @xmath219 ( black ) given by ( [ bnmax ] ) , the envelope @xmath220 ( black ) given by ( [ bnenv ] ) , and the large @xmath0 approximation @xmath221 ( red , dashed ) given by ( [ bnapp]).,title="fig : " ] [ ct][ct][1.5]@xmath288 [ ct][ct][1.5]@xmath192 as a function of @xmath96 for @xmath187 and @xmath144 , respectively . the parameters are @xmath142 , @xmath143 , and @xmath289 .
the blue curve is based on the exact expression given by ( [ 2ldfb2 ] ) , the red plusses are given by the @xmath290 expression in ( [ 2ldfb3]).,title="fig : " ] | we consider the large deviation function for a classical harmonic chain composed of @xmath0 particles driven at the end points by heat reservoirs , first derived in the quantum regime by saito and dhar @xcite and in the classical regime by saito and dhar @xcite and kundu et al .
@xcite . within a langevin description
we perform this calculation on the basis of a standard path integral calculation in fourier space .
the cumulant generating function yielding the large deviation function is given in terms of a transmission green s function and is consistent with the fluctuation theorem .
we find a simple expression for the tails of the heat distribution which turn out to decay exponentially .
we , moreover , consider an extension of a single particle model suggested by derrida and brunet @xcite and discuss the two - particle case .
we also discuss the limit for large @xmath0 and present a closed expression for the cumulant generating function .
finally , we present a derivation of the fluctuation theorem on the basis of a fokker - planck description . this result
is not restricted to the harmonic case but is valid for a general interaction potential between the particles . . |
pulsars are very stable natural clocks with observed steady pulses . however , many pulsars exhibit significant timing irregularities , i.e. , unpredicted arrival times of pulses .
hobbs et al . ( 2010 , hereafter h2010 ) carried out so far the most extensive study of the long - term timing irregularities of 366 pulsars , and ruled out some timing noise models in terms of observational imperfections , random walks ( boynton 1972 ; alpar et al . 1986 ; cheng 1987a , 1987b ) , and planetary companions ( cordes & shannon 2008 ) .
lyne et al . ( 2010 , hereafter l2010 ) found that timing behaviors often result from two different spin - down rates , and pulsars switch abruptly between these states , often quasi - periodically , leading to the observed spin - down patterns .
the deviations of pulsars spin frequency @xmath0 from its long - term trend are correlated with changes in the pulse shapes ( see fig . 3 and 4 of l2010 ) , and therefore are magnetospheric in origin . by modeling the observed precise timing data of pulsars , we will show that the long - term linear change of @xmath1 is consistent with the timescale of hall drift , and the oscillatory structure of @xmath1 and the changes in pulse shapes can be produced by the hall waves in the crust of a neutron star ( ns ) .
consequently the mechanism of magnetic field evolution and the origin of the timing noise of pulsars are better understood , which may improve the sensitivity of gravitational wave detections with pulsars .
assuming the pure magnetic dipole radiation in vacuum as the braking mechanism for a pulsar s spin - down ( e.g. lorimer & kramer 2004 ) , we have @xmath2 in which @xmath3 is a constant , @xmath4 is the strength of the dipole magnetic field at the surface of the ns , @xmath5 , @xmath6 , and @xmath7 are the radius , moment of inertia , and angle of magnetic inclination , respectively .
previously , we constructed a phenomenological model for @xmath4 with a long - term evolution modulated by short - term oscillations ( zhang & xie 2012a and 2012b , hereafter paper i and paper ii , respectively ) : @xmath8 where @xmath9 is the pulsar s age , and @xmath10 , @xmath11 , @xmath12 are the amplitude , phase and period of the @xmath13-th oscillating component , respectively .
@xmath14 , in which @xmath15 is the field strength at age @xmath16 , and @xmath17 is the power law index . then from equation ( [ b evolution ] ) and taking only the dominating oscillating component , we obtained the analytic approximation for @xmath18 ( paper i ) : @xmath19 where @xmath20 is the real age of the pulsar and @xmath21 for the dominating oscillating component .
it has been found that equation ( [ vddot ] ) describes adequately the statistical properties of the observed timing noises of radio pulsars ( paper i ) .
therefore for relatively young pulsars with @xmath22 yr , the first term in equation ( [ vddot ] ) dominates and we should have @xmath23 if @xmath24 . considering that the characteristic ages ( @xmath25 ) of young pulsars are normally several times larger than @xmath26 , equation ( [ vddot ] ) thus explains naturally the observed @xmath27 for most young pulsars with @xmath28 yr .
similarly for much older pulsars , the second term in equation ( [ vddot ] ) dominates , in agreement with the observational fact that the numbers of negative and positive @xmath18 are almost equal for the old pulsars in the sample of h2010 .
the confirmation of equation ( [ vddot ] ) with observations naturally suggests that the observed pulsars indeed have evolutionary links , and the equation therefore provides details about the dipole magnetic field evolution of a young pulsar into an old pulsar .
however , equation ( [ vddot ] ) has so far not been tested with the observed evolutions of individual pulsars .
following the same approach in paper i , we obtain @xmath29 where @xmath30 , @xmath31 describes the long - term monotonic variation of @xmath0
. therefore equation ( [ vdot ] ) can be tested with the long - term monitoring observations of individual pulsars .
we can obtain an estimate of @xmath32 by linearly fitting the long - term monotonic variation of reported @xmath0 .
consequently we can also determine the time scale of the long - term magnetic field evolution of each pulsar ( see equation ( 6 ) in paper i ) @xmath33 for comparisons with theoretical models of neutron star magnetic field evolution .
@xmath34 indicates magnetic field decrease and vice versa .
the sample of l2010 provides the precise histories of @xmath1 for seventeen pulsars and thus may be applied to test equation ( [ vdot ] ) .
lyne et al . ( 2010 ) found that psr b1828@xmath3511 clearly shows a long - term evolution trend ( noticed in l2010 as a linear increase of its @xmath1 ) , i.e. @xmath36 .
we show the comparison between the reported ( taken from l2010 ) and analytically calculated @xmath0 for the pulsar in figure [ fig:1 ] ; all the parameters are listed in table 1 .
the one major difference is caused by the decrease of the oscillation periods of the reported data after @xmath37 days , which indicates that the oscillations are unlikely caused by precession of the neutron star .
nevertheless , our model describes the general trend of the reported data of b1828 quite well . by linearly fitting @xmath1 of all other pulsars
, we find that most of them also exhibit long - term monotonic evolutions modulated by short - term quasi - periodical oscillations .
however , three pulsars in the sample , namely , b1822 - 09 , b2035 + 36 and j2043 + 2740 , exhibit much more erratic short - term behaviors , which may significantly bias the fitting results ; here we exclude these three pulsars from further study . we conclude that our phenomenological model can describe adequately the spin - down evolutions of fourteen pulsars in the sample of seventeen pulsars of l2010 ; all the observed and derived parameters for the fourteen pulsars are listed in table 1 .
we first investigate the alternative possibility that the long - term linear change of @xmath0 is caused by the change of the inclination angle of a ns , rather than the change of its magnetic field strength .
observationally , the inclination angle change would also produce a long - term change on the width of pulses . for a circular beam
, the change of its width can be given by the geometrical relation @xmath38^{1/2},\ ] ] where @xmath39 is the magnitude of the inclination angle change that can be obtained from @xmath40 for psr b1828@xmath3511 ( @xmath41 is the time span of the observations ) , @xmath42 is the impact angle , @xmath43 is the half angular width of the radiating cone ( gil et al .
1984 ) , @xmath44^{\frac{1}{2}},\ ] ] @xmath45 is the initial pulse profile width of b1828@xmath3511 ( l2010 ) .
thus , @xmath46 only depends on @xmath47 and @xmath42 .
there are two observational constraints for @xmath47 and @xmath42 for b1828@xmath3511 : ( a ) @xmath48 ms ( l2010 ) ; and ( b ) its position angle ( p.a . )
gives @xmath49 ( gould & lyne 1998 ) .
we show the comparison of the two constraints on @xmath47 and @xmath42 in figure [ fig:3 ] .
one can see that condition ( a ) requires @xmath50 , which is much smaller than that from condition ( b ) ; thus the possibility of inclination angle change is ruled out for b1828@xmath3511 .
it is therefore reasonable to assume that the long - term linear changes of @xmath0 for other pulsars in the sample are also not caused by the changes of their inclination angles , because the patterns of their spin - down evolutions are very similar to psr b1828@xmath3511 .
goldreich & reisenegger ( 1992 ) studied several avenues for magnetic field decay in isolated nss : ohmic decay , ambipolar diffusion , and hall drift .
depending on the strength of the magnetic fields , each of these processes may dominate the evolution , and the ambipolar diffusion is only important for magnetars .
ohmic decay occurs in both the fluid core and solid crust of a ns ( sang & chanmugam 1987 ; urpin et al .
1994 ; page et al .
it is inversely proportional to the electric conductivity and independent of the strength of magnetic fields . if an electric current flows vertically in magnetic fields , the fields exerts a transverse force on the moving charge carriers .
the resulting hall drift of the carriers can transport magnetic fields from the inner crust to the outer crust ( jones 1988 ) .
the hall effect is non - dissipative and thus can not be a direct cause of magnetic field decay .
however , it can enhance the rate of ohmic dissipation , since only electrons are mobile in the solid crust , and their hall angle is large .
consequently the evolution of magnetic fields resembles that of vorticity , and then the fields undergo a turbulent cascade terminated by ohmic dissipation at small scales ( goldreich & reisenegger 1992 ; muslimov 1994 ; biskamp & m@xmath51ller 1999 ; urpin & shalybkov 1999 ; rheinhardt & geppert 2002 ; geppert & rheinhardt 2002 ; pons & geppert 2010 ; geppert et al .
2013 ) .
cumming et al .
( 2004 , hereafter c2004 ) found that , in isolated nss with relatively pure crusts , the hall effect dominates over ohmic decay after a time @xmath52 , where @xmath53 is in units of @xmath54 g. this is consistent with the fact that @xmath55 is much shorter than the ohmic timescale @xmath56 yr . at lower densities of the crust ,
the degenerated electrons contribute to the pressure , and the hall timescale ( @xmath57 , where @xmath58 is the decay rate induced by hall effect ) is ( c2004 ) @xmath59 in which @xmath60 , @xmath61 is the number fraction of electrons , and @xmath62 is the local gravity , assumed constant . at densities greater than neutron drip ,
the neutrons dominate the pressure ( c2004 ) , @xmath63 where @xmath64 is the number fraction of neutrons , and @xmath65 is the factor that accounts for the interactions between neutrons . in the top panel of figure [ fig:4 ]
, it is shown that there is no significant correlation between @xmath66 and @xmath67 . in the second and third panels ,
we show that @xmath68 is consistent with @xmath67 for eight pulsars and @xmath69 is for the other six pulsars . here , the outer and inner crusts are defined by @xmath70 and @xmath71 , respectively , yielding the upper and lower limits for @xmath72 plotted in the second and third panels of figure [ fig:4 ] .
this implies that the hall drift is responsible for their observed long - term evolutions in @xmath73 . in other words ,
the evolution of their crustal magnetic fields , rather than that in their cores , dominates the observed long term spin - down evolution of these young pulsars . assuming @xmath74
, we obtain an effective density @xmath75 , which may be useful to indicate the location of the majority of the magnetic field lines in the ns crusts .
the values of @xmath75 are shown in the bottom panel of figure [ fig:4 ] .
the hall drift can also pump energy from an internal strong toroidal field to the dipolar poloidal component on a timescale of @xmath76 ( pons et al .
2012 ; gourgouliatos & cumming 2014 ) , resulting in increased @xmath73 for a pulsar . therefore the hall drift can cause both long term decrease and increase of @xmath73 in a pulsar .
figure 3 shows that all the fourteen pulsars are quite young ( with @xmath77 yr ) and only three of the four pulsars are observed with long term increase of @xmath73 , i.e. , @xmath78 .
this is consistent with our previous result that the long - term decrease of magnetic fields are more frequently observed for young pulsars ( paper i ) , suggesting that the hall drift tends to generate more magnetic field decay than increase .
very recently , geppert et al .
( 2013 ) found that the hall drift can produce small - scale strong surface magnetic field anomalies ( spots ) due to the interaction of the initially " dipolar field with the strong toroidal crustal component , which is fundamental for generating observable radio emission .
besides , there is a large scale dipolar poloidal component which has impact on the braking behavior , and its evolution due to hall drift may be responsible for the long - term monotonic evolutions of @xmath1 . the above discussed theoretical expectation that the hall drift can produce
both increase and decrease of @xmath73 suggests that an oscillatory mode of @xmath73 in a ns crust may also be produced ; however , the hall timescales ( either @xmath79 or @xmath80 ) are too long for the observed periodicities in pulsar timing noises ( with @xmath81 yr as listed in table 1 ) . on the other hand , the diffusive motion of the magnetic fields can perturb the background magnetic fields ( here the dipole magnetic fields ) at the base of the ns crust .
such perturbations propagate as circularly polarized
hall waves " along the background field lines upward into the lower density regions in the crusts ( thompson & duncan 1996 ) .
the hall waves can strain the crust with a wave period ( c2004 ) , @xmath82 where @xmath83 is the hall timescale at the base of the crust , and @xmath84 is the number of nodes over the crust .
the elastic response of the crust to the hall wave can induce angular displacement @xmath85 , where @xmath86 is the fluid displacement , @xmath87 is the depth of the crust , and @xmath88 is the shear modulus .
the maximum strain occurs at or above the turning point . using the wkb scaling
, it is found that at the turning point ( c2004 ) , @xmath89 in which @xmath90 is the amplitude of the mode at the base of the crust .
the density at the turning point can then be calculated from ( c2004 ) @xmath91 since most of the dipole field lines are collected at the polar regions of nss , the strains at the regions are especially strong .
the elastic responses of the crusts to the waves induce changes of the areas of polar regions , producing both pulse shape changes and oscillations in @xmath1 .
the quasi - periodical oscillations in @xmath1 and time residuals can be caused by the hall waves , i.e. , @xmath92 .
the calculated parameters from the above equations are also listed in table 1 .
the values of @xmath93 are moderate for the theoretically predicted density at the turning point . above the turning point , the perturbation on magnetic fields @xmath94 decreases to match the boundary condition @xmath95 .
the variation in @xmath96 roughly correlates with @xmath97 as @xmath98 , where @xmath99 is the amplitude of the oscillations in @xmath100 . from equation ( [ vdot ] )
, we have @xmath101 . for b1828@xmath3511 , @xmath102 , which is roughly the same as @xmath103 (
listed in table 1 ) ; here @xmath104 is defined as the peak - to - peak amplitude of the oscillations .
the small difference between @xmath105 and @xmath106 is due to the partial cancellation of the two harmonics with the same amplitude , but different phases . for the other thirteen pulsars ,
@xmath107 is not available and we thus assume @xmath108 , which is reasonable because only one significant periodicity has been observed for each of them . assuming @xmath109 , @xmath110 can be calculated , as listed in table 1 .
finally we can constrain @xmath111 ( the last column in the two tables ) , which is consistent with @xmath112 usually assumed in theoretical calculations ( e.g. thompson & duncan 1996 ) .
in this work , we applied the phenomenological model of pulsar timing noise we developed recently to the observed precise timing data of young radio pulsars , and analyzed the influence of the hall effects in ns crusts on the magnetic field evolutions .
this is the first time that the two aspects of the hall effect , i.e. , hall drift and waves , are correlated with observational data for individual pulsars .
we obtained the following conclusions : 1 .
the @xmath1 evolutions for most of the pulsars in l2010 sample can be described adequately with our phenomenological model consisting a long - term monotonic change and short - term oscillations in @xmath0 ; 2 .
the observed long - term monotonic changes in @xmath0 can be interpreted with the hall drifts in ns crusts ; 3 .
the observed short - term oscillations in @xmath0 and in pulse shapes can be produced by the hall waves in ns crusts . in a unified view , the long - term evolutions in @xmath1 can be understood as that they are caused by the hall waves with very low harmonics ( @xmath113 ) . actually , goldreich & reisenegger ( 1992 ) conjectured a turbulent hall cascade , transferring energy from large to small scales , with an energy spectrum @xmath114 ( where @xmath115 is the wavenumber ) .
however , it is still unclear that why the high harmonics ( e.g. @xmath116 ) dominate in all pulsars we studied here .
it is probably due to some selection biases that the observational time spans for all pulsars are only a few decades .
indeed , hobbs et al .
( 2010 ) found that the structures seen in the timing noise vary with data span ; as more data are collected , more quasi - periodic features are observed .
to ultimately address these questions , the power spectrum and the physical origin of the hall waves need to be investigated in detail in future works .
as we have discussed above and shown previously , the long - term evolution of @xmath73 of young pulsars are quite similar to that of the ( young ) pulsars studied here .
in addition , the huge spans of the reported braking indices of all pulsars can be naturally explained by short - term oscillations of their @xmath73 with characteristics also similar to that of these pulsars .
therefore it is reasonable to assume that the hall effects we studied here are also responsible for the observed spin evolutions of essentially all pulsars , including millisecond radio pulsars .
physically both a long - lived core based dipolar field and a crustal field should coexist and be superimposed for a young pulsar .
however , it is likely that the observed " features , with e.g. @xmath117 yr , are caused by the crustal field , whose evolution proceeds faster than that of the core field , due to hall effect and finite conductivity in its crust . as the pulsar ages , its crustal field is gradually dissipated and
the core field begins to dominate its spin - down evolution . in this scenario
we would expect much longer @xmath118 for millisecond pulsars .
unfortunately , currently the data on the evolutions of @xmath1 of millisecond pulsars are not available with comparable details as for those normal radio pulsars studied here ( l2010 ) .
this is because that millisecond pulsars have much weaker dipole magnetic fields , implying much smaller @xmath110 and @xmath119 , which is actually the basis for using them to detect gravitational waves .
availability of higher quality timing data of millisecond pulsars from the current on - going pulsar timing array observations ( haasteren et al . 2011 ; demorest et al . 2013 ; manchester et al . 2013 ) may allow both the long - term monotonic evolutions and short oscillations to be identified for individual millisecond pulsars , which will improve the current gravitational - wave limits ( shannon et al . 2013 ) .
we thank j. y. liao , j. xu and p. f. wang for useful discussions .
snz acknowledges partial funding support by 973 program of china under grant 2009cb824800 , by the national natural science foundation of china under grant nos .
11133002 , 11373036 and 10725313 , and by the qianren start - up grant 292012312d1117210 .
hobbs , g. , lyne , a. g. , & kramer , m. 2010 , , 402 , 1027 ( h2010 ) boynton , p. e. et al .
1972 , , 175 , 217 alpar , m. a. , nandkumar , r. & pines , d. 1986 , , 311 , 197 cheng , k. s. 1987 , , 321 , 799 ( 1987a ) .
cheng , k. s. 1987 , , 321 , 805 ( 1987b ) .
cordes , j. m. , & shannon , r. m. 2008 , , 682 , 1152 lyne , a. et al .
2010 , science , 329 , 408 ( l2010 ) lorimer , d. r. , & kramer , m. 2004 , handbook of pulsar astronomy , ( cambridge university press , cambridge ) zhang , s. n. , & xie , y. 2012 , , 757 , 153 ( paper i ) zhang , s. n. , & xie , y. 2012 , , 761 , 102 ( paper ii ) gil , j. , gronkowski , p. , & rudnicki , w. 1984 , a&a , 132 , 312 goldreich , p. , & reisenegger , a. 1992 , , 395 , 250 gould , d. m. & lyne , a. g. 1998 , , 301 , 235 gourgouliatos , k. n. & cumming , a. 2014 , mnras , 438 , 1618 sang , y. m. , & chanmugam , g. 1987 , , 323 , l61 urpin , v. a. , chanmugam , g. , & sang , y. m. 1994 , , 433 , 780 page , d. , prakash , n. , lattimer , j. m. , & steiner , a. w. 2000 , phys .
lett . , 85 , 2048 jones , p.b .
1988 , , 233 , 875 muslimov , a. g. 1994 , , 267 , 523 biskamp , d. , & m@xmath51ller , w. c. 1999 , phys .
, 82 , 2195 urpin , v. , & shalybkov , d. 1999 , , 304 , 451 rheinhardt , m. , & geppert , u. 2002 , phys .
lett . , 88 , 101103 geppert , u. , & rheinhardt , m. 2002 , a & a , 392 , 1015 pons , j. a. & geppert , u. 2010 , a&a , 513 , l12 geppert , u. , gil , j. , & melikidze , g. 2013 , , 435 , 3262 cumming , a. , arras , p. , & zweibel , e. 2004 , , 609 , 999 ( c2004 ) pons , j. a. , vigano , d. , & geppert , u. 2012 , a&a , 547 , a9 thompson , c. , & duncan , r. c. 1996 , , 473 , 322 van haasteren , r. , et al .
2011 , , 414 , 3117 demorest , p. b. , et al .
2013 , , 762 , 94 manchester , r. n. , et al .
2013 , pasa , 30 , e017 shannon , r. m. , et al .
2013 , science , 342 , 334 .observed and derived parameters for the pulsars with long - term monotonic changes and short - term quasi - periodical oscillations in @xmath1 ; the former are taken from l2010 . [
cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] _ upper panel _ : @xmath120 for psr b1828@xmath3511 during the past 20 years . the reported data ( taken from l2010 )
are represented by red stars ; and the solid black line is calculated from equation ( [ vdot ] ) with the following parameters : @xmath121 days , @xmath122 , @xmath123 , @xmath124 . the dotted , solid , dot - dashed and dashed lines are linear fits for the major peak values of reported data and analytical calculation , the minor peak values of reported data and analytical calculation , respectively .
_ lower panel _ : time differences between the peak positions of reported data and analytical calculation .
the differences for major peak positions and minor peak positions are represented by triangles and diamonds , respectively.,width=12 ] @xmath66 , @xmath126 , @xmath80 and @xmath75 versus @xmath67 . in all panels : the crosses and circles represent @xmath127 and @xmath128 , respectively . in the second and third panels : the crosses , circles , solid line and dotted line represent @xmath129 , @xmath130 , @xmath131 and @xmath132 , respectively . in bottom panel : the outer crust and inner crust mean the density values that are derived from the equations @xmath131 , and @xmath132 , respectively.,width=18 ] | the observed long - term spin - down evolution of isolated radio pulsars can not be explained by the standard magnetic dipole radiation with a constant braking torque .
however , how and why the torque varies still remains controversial , which is a major issue in understanding neutron stars .
many pulsars have been observed with significant long - term changes of their spin - down rates modulated by quasi - periodic oscillations . applying the phenomenological model of pulsar timing noise we developed recently to the observed precise pulsar timing data of isolated neutron stars , here we show that the observed long - term evolutions of their spin - down rates and quasi - periodic modulations can be explained by hall effects in their crusts .
therefore the evolution of their crustal magnetic fields , rather than that in their cores , dominates the observed long term spin - down evolution of these young pulsars .
understanding of the nature of pulsar timing noise not only reveals the interior physics of neutron stars , but also allows physical modeling of pulsar spin - down and thus improves the sensitivity of gravitational wave detections with pulsars . |
extra dimensional theories @xcite have attracted considerable attention in recent years .
various brane world models provide wide possibilities for phenomenological applications ( for review see , e.g. @xcite ) .
however , the lack of a consistent field theory in extra dimensions compels one to stick to the kaluza - klein approach at the tree level and assume that the string theory cures the problems with divergences at high energy .
one of the immediate consequences of the k - k approach is the increase in the scattering cross section with energy due to the exchange of an infinite tower of k - k modes @xcite . to get a finite result ,
one usually needs to introduce some cutoff , thus making the amplitude essentially cutoff dependent .
the cutoff may well have a physical meaning , for instance , representing the width of the brane , but this does not reduce the amount of arbitrariness . assuming that in the full theory this dependence disappears one can try to look for a qualitative picture . yet
the increase in the total cross section being attractive from the point of view of observations can not last forever due to the breaking of unitarity , so at some energy scale this behaviour has to be changed .
we try to analyze this situation following the conception of the low - energy effective theory based on the wilson approach to the renormalization group @xcite .
though it does not allow one to get a fully consistent theory , it has an advantage of possession of the effective low - energy theory without cutoff and has some remarkable properties at the fixed point @xcite .
we argue that taking into account the radiative corrections in extra dimensions the behaviour of the cross section is changed and finally leads to decreasing dependence on energy .
however , this theory is also valid only up to some energy scale . to illustrate the main features of an effective theory and to compare it with the k - k approach
, we consider below the quark - antiquark scattering due to gluon exchange in the bulk .
assume now that the quark fields are localized on the 3-dimensional brane of zero width while gluons can propagate in the bulk of @xmath3 extra dimensions .
schematically , the situation is shown in fig.[brane ] .
= 2.5 cm the interaction term on the brane looks like @xcite @xmath4 where the coupling @xmath5 , @xmath6 being some characteristic scale ( scale of compactification of extra dimensions , localization scale , etc . ) . at the tree level
the corresponding scattering amplitude in the @xmath7-channel can be written as @xmath8 the euclidean integral over additional momenta in eq.([amp ] ) may be discrete if extra dimensions are compactified
. then one has an infinite sum over the kaluza - klein modes with increasing mass .
if @xmath9 , then the integral ( or a sum ) in eq.([amp ] ) is divergent . usually , one introduces an ultraviolet cutoff @xmath10 which may be understood as inclusion of the nonzero width of the brane or some other way .
this leads to arbitrariness of predictions and depends on unknown details of localization . with the cutoff
the integral takes the form @xcite @xmath11\left ( \frac{s}{m_c^{2 } } \right)^{d/2 - 1}\hspace{-0.3 cm } + \sum_{k=1}^{[(d-1)/2 ] } c_{k } \left ( \frac{s}{\lambda^{2 } } \right)^{k-1 } \left ( \frac{\lambda^2}{m_c^2 } \right)^{d/2 - 1 } \right\},\nonumber\end{aligned}\ ] ] where we have substituted @xmath12 and @xmath13 .
equation([sum ] ) replaces the usual @xmath14 behaviour in 4 dimensions . in the leading order one
has @xmath15 \left ( \frac{s}{m_c^{2 } } \right)^{d/2 - 1}.\ ] ] to glue it to the low energy 4-dimensional theory at @xmath16 , we adjust the scale @xmath17 and finally get @xmath18 \left ( \frac{s}{m_c^{2 } } \right)^{d/2 - 1}.\ ] ] it is easy to see that expression ( [ kkprop ] ) leads to increasing cross - section .
indeed , the differential cross - section , which in the 4-dimensional case looks like @xmath19 now takes the form ( for @xmath20 ) @xmath21 where @xmath22 $ ] and we have taken into account the sign of @xmath23 . in the c.m .
frame one has @xmath24 and , respectively @xmath25 \log^2(\frac{s}{\widetilde{m}_c^2 } ) \right.\nonumber \\ & + & \left .
[ \frac{(1+\cos\theta)^2}{2 } + 2 ] [ \frac{(1-\cos\theta)}{2}]^{d-2 } \log^2(\frac{s(1-\cos\theta)}{2\widetilde{m}_c^2})\right . \label{xdc } \\ & + & \left .
\frac{(1+\cos\theta)^2}{3 } [ \frac{(1-\cos\theta)}{2}]^{d/2 - 1 } \log(\frac{s}{\widetilde{m}_c^2 } ) \log(\frac{s(1-\cos\theta)}{2\widetilde{m}_c^2})\right\ } \left(\frac{s}{m_c^2}\right)^{d } .\nonumber\end{aligned}\ ] ] one can see that the total cross section in the k - k approach increases with energy and has a different angular dependence compared to the 4-dimensional case .
note , however , that eq.([xdc ] ) is true only in the energy region @xmath26 and has to be replaced further by some decreasing function .
we discuss now the low energy effective theory in @xmath27 .
we follow the so - called wilson renormalization group approach @xcite .
consider first the usual gauge theory in @xmath28 dimensions @xmath29.\ ] ] the fields and the coupling have the following canonical dimensions : @xmath30=\frac{d-2}{2}\ , \ \
\ \ [ f]=\frac d2 \ , \ \
\ \ [ g]=2-\frac d2.\ ] ] thus , for @xmath27 the coupling has a negative dimension , which defines some intrinsic scale m , and the theory is nonrenormalizable in the usual sense .
this means that the higher order operators will inevitably be generated in loop expansion . following the wilson approach one has to write down the renormalization group equation which in principle involves an infinite number of operators of type @xmath31 where the coupling @xmath32 is the coupling in the 4+d dimensional theory .
their influence on a full theory is crucial , since they generate an infinite series of counterterms with arbitrary coefficients . not pretending to resolve this problem at high energy we go now in the infrared direction ,
i.e. , take @xmath33 .
then the situation is somewhat simplified .
one finds that all the higher dimensional operators are irrelevant being suppressed by powers of @xmath34 and may be ignored at the tree level @xcite .
the only gauge invariant operator which is left is the one of the lowest dimension , namely , @xmath35 .
operator in case of a scalar theory of critical phenomena @xcite . ]
when going to the loop expansion the situation is more tricky .
formally , all the uv divergences are absorbed into the redefinition of the higher order operators which are irrelevant at low energy .
as for the finite contributions of these operators to the green functions one expects that they are also suppressed compared to the leading contribution from the lowest dimensional operator @xcite . having all these in mind and
leaving a single operator @xmath35 one can write down the rg equation for the numerical coupling .
since it is dimensionful it is useful to take the dimensionless combination @xmath36 = 0,\ ] ] where @xmath17 is some scale . in what follows
we proceed according to the so - called @xmath37-expansion procedure and consider a theory in @xmath38 dimensions performing renormalization in the vicinity of the critical dimension @xmath39 .
then the rg equation has a simple form @xmath40 where the dots stand for all those irrelevant operators that we ignore at low energy . here
@xmath41 is the usual anomalous dimension calculated within some renormalization scheme . in general @xmath42 , and
hence @xmath43 , may depend on @xmath28 being finite while @xmath44 .
however , in the ms - scheme this dependence is absent and @xmath41 can be calculated directly in the critical dimension @xmath39 .
equation([rg ] ) has a nontrivial fixed point @xmath45 .
it is the so - called wilson - fisher fixed point @xcite usual in scalar theories for @xmath46 .
however , since in gauge theories , contrary to the scalar case , the beta function in critical dimension is negative , the fixed point of this kind exists for @xmath27 @xcite .
it is ultraviolet stable , the coupling approaches it when momentum increases . at the fixed point
the theory possesses no scale : the effective coupling becomes dimensionless @xcite . though the value of the coupling is unknown , the value of the anomalous dimension is known _ exactly _ :
it is not small but is an integer and includes all the radiative corrections . in the vicinity of the fixed point
the green functions exhibit a power - like behaviour .
in particular the gluon propagator in @xmath28 dimensions behaves like @xmath48 in the background gauge , which is appropriate for our case since the beta function here is completely defined by the propagator , @xmath49 .
the latter can be calculated by perturbation theory or found from eq.([rg ] ) at the fixed point . at the same time , the low energy effective theory may be used in the limited energy interval @xmath50 , where @xmath6 is the compactification scale and @xmath51 is the intrinsic scale of a higher dimensional theory . for higher energies
it should be replaced by a full theory .
consider now the cross - section of the quark - antiquark scattering in the et approach .
it is given by the same diagrams , but now one has to use the modified gluon propagator due to eq.([pr ] ) . at high energy the behaviour of the gluon propagator is affected by the fixed point since it is uv attractive .
however , here we have a kind of a controversy : our et is a low - energy one since we ignore all the higher order operators and we move in the uv direction where their influence is essential .
not being able to go beyond the scale @xmath52 within the et approach we can still solve the rg equations in the vicinity of the fixed point . if the initial value of the coupling is close enough to its fixed point value , the influence of the fixed point is essential even at finite energy .
we proceed in two steps .
first we consider the coupling equal to its fixed point value .
the advantage is that the anomalous dimension of the gluon propagator is known exactly to all orders of perturbation theory and is equal to @xmath53 .
then the integral ( [ amp ] ) becomes @xmath54 it is convergent now for any @xmath3 and equals @xmath55 note that for @xmath56 we have the standard answer @xmath57 . at the fixed point @xmath58
and we have @xmath59 this means that we have the same behaviour of the cross section as in 4 dimensions !
the only difference is the coefficient that depends on the number of extra dimensions and the coupling at the fixed point .
the latter is unknown but can be calculated in perturbation theory .
the result is not surprising since it is given essentially by dimensional analysis : at the fixed point the theory is scale invariant and a naive power counting is valid .
consider now the initial conditions when the coupling is smaller but close to the fixed point .
we have no all loop information now , but bearing in mind that we are still in perturbative regime , we take the one - loop approximation .
one has @xmath60 then the solutions of the rg equations are @xmath61 , \\
g_a(p^2 ) & = & \frac{1}{p^2 } \frac{\displaystyle 1}{\displaystyle 1+\frac{2b}{d}\tilde g_0 ^
2 \left[\left(\frac{p^2}{p_0 ^ 2}\right)^{d/2}-1\right ] } , \label{fg}\end{aligned}\ ] ] where @xmath62 is a @xmath63-dimensional momentum .
replacing now the propagator in eq.([amp ] ) by ( [ fg ] ) one gets @xmath64}\ .\ ] ] this integral is again convergent for any @xmath3 . to get an explicit expression we calculate it for @xmath65 .
the result is @xmath66 in the limit @xmath67 one gets @xmath68 which coincides with ( [ fpr ] ) for @xmath69 and @xmath65 . to link with the low - energy 4-dimensional theory , we normalize it to @xmath14 at @xmath16
this gives the equation @xmath70 this is a single equation for two dimensionless variables : @xmath71 and @xmath72 .
taking , for instance , @xmath73 ( this choice corresponds to qcd with 6 flavours and @xmath74 ) , we get the equation @xmath75 the solution is @xmath76 .
the same propagator has to be used in the @xmath23-channel , but again one has to take into account the sign change . to get the cross - section , one has to replace the propagators in eq.([x4 ] ) by their expressions in the effective theory ( [ ex ] ) when the argument exceeds @xmath77 .
the fixed - point cross section is obtained by taking the asymptotic form of the propagator ( [ fpr ] ) . as an illustration
, we show in fig.[cross ] the resulting cross section in the case of @xmath65 ( or @xmath78 ) for @xmath79 .
the compactification scale @xmath6 is taken to be equal to 1 tev and the parameters @xmath80 and @xmath81 as above .
for comparison we show also the 4-dimensional cross section ( [ x4c ] ) and the k - k one ( [ xdc ] ) .
one can see that the cross section in the effective theory decreases with energy in agreement with unitarity and in contradiction with the k - k approach .
it slowly departs from the 4-dimensional cross section starting from @xmath82 by changing the slope and finally approaches the fixed - point solution .
note , however , the range of validity of effective theory .
in should be emphasized that the problem we discuss has no solid background due to nonrenormalizability of original higher dimensional theory .
therefore , any attempts to get the low - energy results either within the k - k approach or the et one inevitably faces the problem of arbitrariness while handling with the uv behaviour .
the naive cutoff is not a solution to this problem even if irrelevant operators are suppressed at low energy because the results for various processes essentially depend on the cutoff procedure . from this point of view the effective theory approach based on a fixed point within the wilsonian renormalization group
has an advantage of being universal at low energy .
one can see that the cross - section calculated within the et approach essentially departs from the one calculated in the leading order of the k - k approach .
the reason is that the radiative corrections are now power - like and are not the logs as in 4-dimensional theory .
therefore , their summation drastically changes the behaviour for increasing momenta .
of course , our curves have sense only in the ir region ; however , as we have already mentioned , in the vicinity of the fixed point even for small momenta the behaviour is governed by its attraction .
we considered quark - antiquark scattering via gluon exchange as an example to probe the theory at higher dimensions .
apparently the same situation takes place for other reactions and other exchange quanta .
this means that the increase in the cross - section with energy which was considered to be the `` smoking gun '' for extra dimensional theory , probably is not the case .
based on our formulas we conclude that the manifestation of possible extra dimensions is the change of the slope with modified angular dependence , though the signal is not pronounced and has to be studied in more detail for a particular process .
we are grateful to v.rubakov , a.slavnov , y.okada , and o.teryaev for useful discussions . financial support from rfbr grant # 02 - 02 - 16889 and grant of the ministry of science and technology policy of the russian federation # 2339.2003.2
is kindly acknowledged .
d.k.i would like to thank the theory group of kek where this paper was finished for support and hospitality .
99 n.arkani-hamed , s.dimopoulos , and g.dvali , phys.lett .
* b429 * ( 1998 ) 263 ; phys.rev . *
d59 * ( 1999 ) 086004 ; + l.randall and r.sundrum , phys.rev.lett .
* 83 * ( 1999 ) 3370 ; ibid 4690 .
yu.a.kubushin , models with extra dimensions and their phenomenology , + hep - ph/0111027 ; + j.hewett and m.spiropulu , particle physics probes of extra spacetime dimensions , hep - ph/0205106 .
g.f.giudice , r.rattazzi and j.d.wells , nucl.phys . *
b544 * ( 1999 ) 3 .
k.g.wilson , phys.rev . *
b4 * ( 1971 ) 3174 , 3184 ; + k.g.wilson and j.b.kogut , phys.rep . *
12c * ( 1973 ) 75 ; + k.g.wilson ans m.e.fisher , phys.rev.lett . *
28 * ( 1972 ) 240 .
d.kazakov , jhep 0303 ( 2003 ) 020 .
d.kazakov , field theory in extra dimensions , hep - th/0311211 .
s.weinberg , `` the quantum theory of fields '' .
vol . 1 , ch.12.3 , cambridge , uk : univ
. press , 2000 .
m.peskin , phys.lett . *
94b * ( 1980 ) 161 . | we analyze the behaviour of the high - energy scattering amplitude within the brane world scenario in extra dimensions .
we argue that contrary to the popular opinion based on the kaluza - klein approach , the cross - section does not increase with energy , but changes the slope close to the compactification scale and then decreases like in the 4-dimensional theory .
a particular example of the quark - antiquark scattering due to the gluon exchange in the bulk is considered . _
@xmath0bogoliubov laboratory of theoretical physics , joint institute for nuclear research , dubna , russia + @xmath1institute for theoretical and experimental physics , moscow , russia + @xmath2moscow institute of physics and technology , moscow , russia _ |
weakly interacting massive particles ( wimps ) , and in particular the lightest neutralino in supersymmetric models , are a well motivated dark matter candidate @xcite .
wimps can be detected directly , in the lab , via the elastic scattering of wimps on detector nuclei @xcite .
experiments now have the sensitivity required to probe the theoretically favoured regions of parameter space @xcite and the cdms experiment has recently observed two events in its wimp signal region @xcite .
neutrons , from cosmic - ray induced muons or natural radioactivity , can produce nuclear recoils which ( on an event by event basis ) are indistinguishable from wimp induced recoils .
furthermore perfect rejection of other backgrounds is impossible , for instance ` surface events ' ( electron recoils close to the detector surface ) in the case of cdms .
as highlighted by the cdms events , as well as the long - standing dama annual modulation signal @xcite , demonstrating the wimp origin of a putative signal is absolutely crucial .
the direction dependence of the wimp scattering rate ( due to the earth s motion with respect to the galactic rest frame ) @xcite provides a potential wimp ` smoking gun ' . assuming that the wimp distribution is predominantly smooth ,
the peak wimp flux comes from the direction of solar motion ( towards the constellation cygnus ) and the recoil rate is then peaked in the opposite direction .
the recoil rate , in the galactic rest frame , is highly anisotropic ; the rate in the forward direction is roughly an order of magnitude larger than that in the backward direction .
a detector capable of measuring the nuclear recoil vectors ( including the sense + @xmath1 versus -@xmath2 ) in 3-dimensions , with good angular resolution , could reject isotropy of the recoils with only of order 10 events @xcite .
most , but not all , backgrounds would produce an isotropic galactic recoil distribution ( due to the complicated motion of the earth with respect to the galactic rest frame ) . an anisotropic galactic recoil distribution would therefore provide strong , but not conclusive , evidence for a galactic origin of the recoils .
confirmation of the wimp origin could be obtained by verifying that the properties of the anisotropy match the expectations for wimp induced recoils . assuming the wimp distribution is dominated by a smooth component , the median inverse recoil direction
should be compatible with the direction of solar motion . to summarise , a wimp search strategy with a directional detector
could be divided into a sequence of phases : + 1 .
search phase ( detection of non - zero recoil signal ) + 2 .
detection of anisotropy + 3 .
study of properties of anisotropy + which require successively larger numbers of events ( and hence larger exposures ) .
the initial simple search phase aims to detect an anomalous recoil signal above that expected from backgrounds .
to claim an anomalous signal inconsistent with zero at 95% confidence requires 4 - 5 events .
the second step , as discussed above , would be to check whether the galactic recoil directions are anisotropic and would require of order 10 events ( assuming zero background ) .
in this paper we focus on the third phase , examining how measuring the median recoil direction could be used to provide confirmation of the wimp origin of an anisotropic recoil signal . in section [ model ]
we describe the input to our monte carlo simulations . in section [ res ]
we present our results , before concluding with discussion in sec .
we use the same statistical techniques and methods for calculating the directional nuclear recoil spectrum as in refs . @xcite .
we briefly summarise these procedures here , for further details see these references and ref .
@xcite .
many of the directional detectors currently under development @xcite are low pressure gas time projection chambers ( tpcs ) , e.g. dmtpc @xcite , drift @xcite and newage @xcite .
we therefore simulate a fairly generic tpc based detector .
we assume that the the recoil directions , including their senses , are reconstructed perfectly in 3d .
these are optimistic assumptions , therefore our results provide a lower limit on the number of events / exposure required by a real tpc based detector . for concreteness
we use a @xmath3 target with an energy threshold of 20 kev .
finite angular resolution does not significantly affect the number of events required to detect anisotropy @xcite , provided it is not worse than of order @xmath4 .
axial and/or 2-d read - out would , however , significantly degrade the detector capability @xcite .
the detailed angular dependence of the recoil rate depends on the exact form of the wimp velocity distribution @xcite .
however , if the wimp velocity distribution is dominated by a smooth component the main features of the recoil distribution ( rear - front asymmetry , median direction opposite to the direction of solar motion ) are robust ( see e.g. ref .
@xcite ) ) . for concreteness
we use the standard halo model halo , an isotropic sphere with local density @xmath5 and a maxwellian / gaussian velocity distribution with three dimensional velocity dispersion @xmath6 , and fix the wimp mass at @xmath7 .
numerical simulations find velocity distributions with stochastic features at large speeds @xcite .
kuhlen et .
@xcite find that for high speed wimps ( @xmath8 ) the direction in which the wimp flux is largest can deviate by more than @xmath9 from the direction of solar motion .
the effect on the median recoil direction will be substantially smaller than this however .
firstly the set - up we are considering ( @xmath3 target with energy threshold of 20 kev and @xmath7 ) is sensitive to wimps with @xmath10 .
therefore a high speed feature will only contribute a small fraction of the wimp flux .
this will be true in general unless the wimp mass is small and/or the threshold energy is large . secondly , see e.g. fig . 3 of ref .
@xcite , due to the elastic scattering the recoil rate is less anisotropic than the wimp flux .
the deviation of the inverse median direction from the direction of solar motion will therefore be substantially smaller than the deviation of the peak wimp flux .
it will also depend on the ( a priori unknown ) velocity and density of the feature .
for instance , for a stream with velocity , in galactic coordinates , @xmath11 the difference between the inverse median direction and the solar direction only exceeds @xmath12 if the fraction of the local density in the stream exceeds @xmath13 @xcite . since the effect of a feature in the speed distribution on the median recoil direction
is expected to be small we do not investigate it in this study . if , with a large number of events , a statistically significant deviation of the inverse median direction were found , its origin could be investigated by studying the energy dependence of the deviation .
we defer an investigation of this to future work . finally , direct detection experiments probe the ultra - local dark matter distribution on scales many orders of magnitude smaller than those resolved by simulations .
it is not ( and may never be ) possible to directly calculate , or otherwise measure , the dark matter distribution on the relevant scales . in this case
if wimps are detected , then the directional recoil rate could be used to reconstruct the ultra - local dark matter distribution @xcite .
the recoil rate is peaked in the direction opposite to the direction of solar motion . to allow ease of comparison with the direction of solar motion we use the inverse recoil directions ( i.e. the directions from which the recoils originate ) in our analyses .
we first examine how the accuracy with which the median galactic recoil direction is determined depends on the number of wimp events .
the median direction is defined as the direction @xmath14 which minimises the sum of the arclengths between @xmath15 and the individual inverse recoil directions @xmath16
it is found by minimising @xmath17 where @xmath18 is the number of events . for a given number of wimp events , @xmath19
, we simulate @xmath20 experiments and in each determine the direction , @xmath15 , by minimizing eq .
( [ m ] ) and hence calculate @xmath21 , @xmath22 the angle between the median direction and the direction of solar motion , @xmath23 . in fig .
1 we plot the 50% and 95% percentiles of the distribution of @xmath21 as a function of @xmath24 .
we also investigate the effects of non - zero ( isotropic ) backgrounds .
we parametrize the background rate in terms of the fraction of events which are signal , @xmath25 where @xmath3 and @xmath26 are the signal and background rates respectively ( c.f .
@xcite ) . for comparison
we also plot the @xmath27 and @xmath28 percentiles for an purely isotropic distribution .
, as a function of the number of wimp events , @xmath19 , for varying signal fractions ( from top to bottom ) @xmath29 and @xmath30 .
the solid and dashed lines are the @xmath31 and @xmath28 percentiles .
the dotted lines are ( from top to bottom ) the @xmath28 and @xmath27 percentiles for an isotropic recoil distribution.,width=321 ] we now determine the number of events required to confirm the direction of solar motion as the median inverse recoil direction at 95% confidence .
we do this using the same methodology as in ref .
briefly , we use the distribution of @xmath21 for wimp induced recoils and for the null hypothesis of isotropic recoils , to calculate the rejection and acceptance factors , @xmath32 and @xmath33 .
the rejection factor gives the confidence level with which the null hypothesis can be rejected given a particular value of the test statistic , while the acceptance factor is the probability of measuring a larger value of the test statistic if the alternative hypothesis is true , we then find the number of wimp events required to give @xmath34 i.e. to reject the median direction being random at 95@xmath35 confidence in @xmath31 of experiments .
this is shown in fig .
[ nlambda ] , as a function of the signal fraction @xmath36 . for zero background ( i.e. @xmath37 )
31 events are required , a factor of @xmath0 more than are required to simply reject isotropy .
the number of events required increases as @xmath36 is decreased , initially fairly gradually and then , once the signal becomes subdominant ( @xmath38 ) , more rapidly .
billard et al .
@xcite have recently investigated using a sky - map based likelihood analysis to probe the correlation of the directional recoil rate with the direction of solar motion ( and hence confirm the wimp origin of a signal ) .
their results are qualitatively consistent with ours .
for instance they find that , for a signal fraction @xmath39 , the peak signal direction can be confirmed to be within @xmath40 of the direction of solar motion with as few as 25 wimp events .
the median direction method is however faster and more robust .
it particular it has the advantage of being model independent ( i.e. one does not need to assume an exact form of the wimp velocity distribution ) . , required to reject the median direction being random at 95@xmath35 confidence in @xmath31 of experiments on the signal fraction , @xmath36.,width=321 ]
if desired , further confirmation of the wimp origin of an anisotropic recoil signal could be obtained by studying the variation of the median recoil direction in the lab frame @xcite . over the course of a sidereal day the peak in the recoil distribution traces out a small circle on the sky . for a detector located in the northern hemisphere during a sidereal day
the peak recoil direction in the lab rotates ( roughly ) from down to south and back .
a periodogram analysis could be used to verify that median direction varies over a sidereal , rather than solar , day ( c.f .
@xcite for the annual modulation ) .
in this paper we have investigated how the median recoil direction in directional detection experiments can be used as a wimp signal . assuming a smooth wimp distribution , the peak wimp flux is from the direction of solar motion and the median recoil direction is in the opposite direction .
an ideal detector could reject isotropy of recoils with only of order 10 events @xcite .
confirmation of the wimp origin of an anisotropic recoil signal could be obtained by studying the details of the anisotropy and in particularly confirming that the median inverse galactic recoil direction coincides with the direction of solar motion .
we find that , with an ideal detector and zero - background , to confirm the direction of solar motion as the median inverse recoil direction at 95% confidence requires 31 events ( see also ref .
non - zero isotropic background would increase this number , significantly if the signal is sub - dominant . for concreteness ( and in the absence of a definitive alternative ) we have used the standard halo model halo , which has an isotropic maxwellian speed distribution . while the detailed angular dependence of the recoil rate depends on the exact form of the wimp velocity distribution ,
if the wimp velocity distribution is dominated by a smooth component the median inverse recoil distribution will be close to the direction of solar motion ( see e.g. ref .
recent high resolution simulations have found stochastic , features in the speed distribution at large speeds @xcite .
the direction in which the flux of high speed wimps is largest can deviate by more than @xmath9 from the direction of solar motion @xcite .
however the deviation of the median inverse direction will be small compared with that expected from an isotropic recoil direction .
we therefore conclude that features in the speed distribution at high speed are unlikely to affect the utility of the median recoil direction as a wimp signal . if , with a large number of events , a statistically significant deviation of the inverse median direction were found , its origin could then be investigated by studying the energy dependence of the deviation . | direct detection experiments have reached the sensitivity to detect dark matter wimps . demonstrating that a putative signal is due to wimps , and not backgrounds , is a major challenge however .
the direction dependence of the wimp scattering rate provides a potential wimp ` smoking gun ' .
if the wimp distribution is predominantly smooth , the galactic recoil distribution is peaked in the direction opposite to the direction of solar motion .
previous studies have found that , for an ideal detector , of order 10 wimp events would be sufficient to reject isotropy , and rule out an isotropic background .
we examine how the median recoil direction could be used to confirm the wimp origin of an anisotropic recoil signal .
specifically we determine the number of events required to confirm the direction of solar motion as the median inverse recoil direction at 95% confidence .
we find that for zero background 31 events are required , a factor of @xmath0 more than are required to simply reject isotropy .
we also investigate the effect of a non - zero isotropic background . as the background rate
is increased the number of events required increases , initially fairly gradually and then more rapidly , once the signal becomes subdominant .
we also discuss the effect of features in the speed distribution at large speeds , as found in recent high resolution simulations , on the median recoil direction . |
distinct surface - electronic properties , potentially relevant for spintronic applications , arise from the strong spin - orbit interaction in three - dimensional topological insulators ( 3dtis ) @xcite .
although the bulk electronic structure of these materials resembles that of standard band insulators with electronic bands separated by an energy gap , the valence and conduction bands of the surface states form a conical dispersion and touch at the center of the surface brillouin zone .
these gapless surface states lack the standard twofold spin degeneracy , they are protected against backscattering , and the spin orientation of each plane - wave surface state is determined unambiguously by its momentum vector . in the past few years , surface - sensitive experimental techniques
have been utilized to explore the remarkable properties of the surface electrons in 3dtis .
the linear , dirac - cone - like electronic dispersion and deviations from that were observed in various 3dti materials using angle - resolved photoemission spectroscopy @xcite ( arpes ) , and the correlation between spin and momentum was demonstrated by the spin - resolved version of the same technique @xcite .
the role of electron scattering off pointlike impurities and line defects on 3dti surfaces , highly relevant for future attempts to design electronic devices based on these materials , has been studied via scanning tunneling microscopy ( stm ) @xcite . in the vicinity of obstacles on the surface ,
characteristic standing wave patterns are formed due to the interference of initial and final scattering states @xcite .
these electronic standing waves contribute to the local density of states ( local dos , ldos ) , therefore real - space mapping of them is possible via stm .
theories describing the standing waves on 3dti surfaces have also been formulated recently @xcite .
a line defect has translational symmetry in the direction it stretches along , hence the electronic standing waves in its vicinity are essentially one - dimensional ( 1d ) , ( i.e. , the ldos varies only along the axis perpendicular to the line defect ) .
this simple 1d character of the induced ldos pattern implies a relatively straightforward experimental and theoretical analysis of the effect , which serves as a strong motivation to consider such arrangements .
a line defect arises naturally at the edge of a step formed by an extra crystal layer on the surface , @xcite hence this 1d setup is accessible experimentally .
information on the electronic system can be extracted from the asymptotic decay exponent and wave number of ldos oscillation around line defects .
theoretical results @xcite indicate that the ldos oscillations on the surface of a 3dti , asymptotically far from a line defect and within the energy range of linear dispersion , decay with the distance @xmath2 from the defect as @xmath3 .
this decay exponent is in contrast with the @xmath4 decay seen in a standard two - dimensional electron gas , @xcite and arises as a consequence of the absence of backscattering characteristic of surface electrons in 3dtis .
recent stm data agrees with this prediction .
@xcite the wave vector of the asymptotic ldos oscillations is usually equal to the distance between nesting segments of the constant - energy contour ( cec ) , which is the diameter of the fermi circle in the above - mentioned case .
this has been used in a recent experiment @xcite to confirm the linear dispersion and to infer the fermi velocity on the dirac cone in bi@xmath0te@xmath1 . for energies well above the dirac point ,
the topological surface conduction band of bi@xmath0te@xmath1 is subject to strong hexagonal warping .
stm data corresponding to this energy range is available @xcite , however , the rather complex geometry of the dispersion relation has so far prevented an unambiguous theoretical interpretation of the observations . in this work ,
we provide a theoretical investigation of ldos oscillations created by a line defect on the surface of bi@xmath0te@xmath1 .
we describe the effect in an exact scattering - theory framework @xcite , yielding results that are not restricted to the spatial region asymptotically far from the defect , but hold also in the vicinity thereof .
this enables us to directly compare our results with experimental data , the latter being usually taken close to the defect where features of the ldos are most pronounced . in the energy range of strong hexagonal warping , we identify a significant pre - asymptotic contribution to the ldos oscillations , with wave number quantitatively matching that of a recent experiment @xcite .
te@xmath1 along the @xmath5 ( solid line ) and @xmath6 ( dashed line ) directions of the surface brillouin zone .
( c , d ) surface - state conduction - band density of states of bi@xmath0te@xmath1 . for ( a ) and ( c ) , eq . was used with parameter values@xcite @xmath7 ev , @xmath8 ev@xmath9 and @xmath10 . for ( b ) and
( d ) , the parameter values @xmath11 , @xmath12 , @xmath13 and @xmath14 were used .
red crosses ( blue points ) represent the measured dispersion along @xmath5 ( @xmath6 ) ( data taken from ref . ) .
the zero of energy corresponds to the dirac point of the spectrum .
energy intervals overlapping with the bulk valence band ( bvb ) and bulk conduction band ( bcb ) are also shown.,scaledwidth=48.0% ]
in order to base our forthcoming calculations on an accurate surface - band dispersion , we first establish accurate values of the relevant band - structure parameters ( defined below ) of bi@xmath0te@xmath1 .
arpes measurements @xcite indicate that the surface bands of 3dtis with the crystal structure of bi@xmath0te@xmath1 are subject to hexagonal warping , which can be described by the envelope - function hamiltonian @xcite : @xmath15 where @xmath16 , and @xmath17 . here
, @xmath18 is the vector of pauli matrices representing spin , @xmath19 and @xmath20 are band - structure parameters , @xmath21 , and @xmath22 and @xmath23 are momentum components along the @xmath5 and @xmath6 directions of the surface brillouin zone , respectively . for convenience , we performed a @xmath24 rotation around the @xmath25 axis compared to the hamiltonian in ref . .
energy eigenvalues of @xmath26 in eq .
are @xmath27 where @xmath28 ( @xmath29 ) stands for conduction ( valence ) band .
higher - order terms@xcite in @xmath30 can also be included in @xmath26 . satisfactory agreement between the spectrum in eq . and the arpes spectra of bi@xmath0se@xmath1 surface states @xcite can be obtained by neglecting the band - structure parameters @xmath31 and @xmath20 . in bi@xmath0te@xmath1
however , the conduction - band dispersion measured along the @xmath5 direction , shown with red crosses in figs . [
fig : dos]a , b , has a sub - linear segment , which can be theoretically reproduced by eq . only if @xmath31 and @xmath20 are finite .
we find that using the band - structure parameter set @xmath11 , @xmath12 , @xmath13 and @xmath14 , the measured dispersion relations along the @xmath6 and @xmath5 directions and the surface density of states ( fig .
1c , d of ref . )
are accurately described by @xmath26 [ eq .
] up to 335 mev above the dirac point .
we use these parameter values throughout this paper .
[ fig : dos]a and [ fig : dos]b ( fig .
[ fig : dos]c and [ fig : dos]d ) compares the theoretical surface - state dispersion ( density of states ) calculated with a parameter set used in an earlier work@xcite , and the above parameter set that we found to be optimal , respectively .
further considerations used to find the optimal parameter set above are included in appendix [ sec : parameters_appendix ] .
our goal is to theoretically describe the oscillations in the surface - state conduction - band ldos induced by a line defect , e.g. , the edge of an atomic terrace @xcite , on the surface of bi@xmath0te@xmath1 . for the moment
we assume that the defect forms a straight line that coincides with the @xmath32 axis ( see fig .
[ fig : setup ] ) . following ref . , we model the system with the hamiltonian @xmath33 , where effect of the defect is described via the potential @xmath34 .
region ( solid arrow ) is partially reflected ( dashed arrow ) and transmitted ( dotted arrow ) at a line defect , e.g. , an atomic terrace , on the surface of bi@xmath0te@xmath1 . ( b )
hexagonally warped constant - energy contour in reciprocal space , corresponding to energy @xmath35 mev above the dirac point .
incident and reflected wave vectors from ( a ) are also shown.,scaledwidth=40.0% ] our analysis of the ldos oscillations is based on exact energy eigenstates describing scattering of conduction - band electrons by the line defect ( see fig .
[ fig : setup ] ) .
therefore we first describe the scattering process of a plane wave energy eigenstate @xmath36 incident from , say , the @xmath37 side of the edge , with momentum components @xmath38 and @xmath39 , energy @xmath40 , and spin wave function @xmath41 .
scattering at the line defect is elastic , hence the energy @xmath40 is conserved .
the momentum component @xmath42 parallel to the defect is also conserved due to translational invariance in the @xmath32 direction .
the value of @xmath42 determines the number of propagating waves at a given energy .
for example , in fig .
[ fig : setup]b , the number of propagating waves can be two ( i and r ) or four ( i@xmath43 , i@xmath0 , r@xmath43 , r@xmath0 ) , depending on @xmath42
. however , the incident plane wave can be scattered into coherent superpositions of three reflected and three transmitted partial waves , for the following reasons . on the @xmath44 side of the defect ,
the equation @xmath45 has six complex solutions @xmath46 for given values of @xmath40 and @xmath42 , which follows from eq .
( the numerical method used to obtain these solutions is described in appendix [ sec : planewave ] ) .
three of the @xmath47-s correspond to propagating waves moving away from the defect or evanescent modes .
the associated wave functions @xmath48 ( @xmath49 ) should be included in the ansatz of the complete scattering state .
the remaining three solutions @xmath47 @xmath50 correspond to propagating waves towards the defect or diverging modes , hence they are disregarded .
these arguments , together with their generalization to transmitted waves , imply that the @xmath2-dependent component of the complete scattering wave function is @xmath51 where @xmath52 here , the @xmath53-s and @xmath54-s are reflection and transmission coefficients , @xmath55 is the @xmath2-component of the group velocity of the plane wave with wave - vector components @xmath56 , and the factors @xmath20 and @xmath57 ensure the unitarity of the scattering matrix built up from the reflection and transmission coefficients .
evanescent modes are not subject to the unitarity requirement , hence we are allowed to make the above arbitrary choice @xmath58 for partial waves with complex wave vectors . as the hamiltonian is a third - order differential operator , partial waves at the two sides of the defect should be matched via boundary conditions ensuring their continuity as well as the continuity of their first and second derivatives .
the scattering state @xmath59 of a plane wave incident from the left side ( @xmath60 ) of the defect can be described analogously .
the ldos at a given energy @xmath40 and position @xmath2 is expressed with the exact scattering states as @xmath61 where @xmath62 [ @xmath63 is the wave - vector contour of waves that ( i ) are incident from the right [ left ] side of the line defect , and ( ii ) have energy @xmath40 .
note that @xmath64 breaks up to disconnected pieces for strong hexagonal warping , e.g. , in fig .
[ fig : setup]b , the points @xmath65 and @xmath66 belong to @xmath62 but @xmath67 and @xmath68 do not .
( the wave - vector contour @xmath62 is shown in fig .
[ fig : results]b , there it is formed as the union of the thick blue lines . ) the infinitesimal line segment along @xmath69 is denoted by @xmath70 , and @xmath71 is magnitude of the group velocity of the wave with momentum vector @xmath56 . using the exact scattering wave functions @xmath72
, we evaluate the integral in eq . numerically . to account for the inevitable roughness of the line defect and to suppress noise due to the limited precision of the numerical integration
, we average the ldos oscillations @xmath73 over the angular range @xmath74 $ ] of the line defect orientation with respect to the @xmath6 direction .
in fig . [ fig : results]a , we plot the numerically computed ldos oscillations @xmath75 ( black points ) on the @xmath44 side of the defect , corresponding to energy @xmath35 mev and potential step height @xmath76 mev .
recent theories using asymptotic analysis @xcite suggest that the ldos oscillations in the vicinity of a line defect on the surface of bi@xmath0te@xmath1 decay no faster than @xmath3 .
motivated by this finding , we fit the function @xmath77 via fitting parameters @xmath78 , @xmath79 , and @xmath80 to the results ( shown as red solid line ) .
we also fit an exponentially decaying function @xmath81 via fitting parameters @xmath82 , @xmath83 , @xmath84 and @xmath85 ( blue dashed line ) .
the two major features of our numerical result @xmath73 are as follows .
( i ) comparison of the three curves suggests that the decay of @xmath73 is better described by the exponentially decaying function @xmath86 than by @xmath87 having power - law decay ( see appendix [ sec : fourier ] for further details ) .
( ii ) the wave - number value obtained from fitting @xmath86 is @xmath88 @xmath89 .
expectations for the wave number of the ldos oscillations can be drawn from asymptotic analysis @xcite .
that suggests that the wave number of an electronic standing wave at a given energy @xmath40 is associated to wave vectors connecting nesting segments of the corresponding cec @xcite , i.e. , @xmath90 @xmath89 or @xmath91 @xmath89 depicted in fig .
[ fig : results]b .
as the wave number @xmath92 characteristic of our data @xmath93 deviates significantly from @xmath94 and @xmath95 , and its decay is exponential rather than power - law , we conclude that @xmath93 is dominated by a pre - asymptotic contribution in the considered spatial range . in what follows
, we argue that ( i ) the pre - asymptotic component of the ldos is due to the interference of incoming and reflected partial waves , i.e. , the role of evanescent and transmitted partial waves is negligible , and ( ii ) the appearance of the characteristic wave number @xmath92 in the ldos oscillations is due to the non - monotonic behavior of the parallel - to - defect group velocity component along the cec . to this end , we now consider only the interference contribution of incoming and reflected propagating waves to the ldos [ eq . ] on the right half plane @xmath96 , rewrite it as an integral over the perpendicular - to - defect wave - vector component @xmath97 , and drop the contributions from @xmath97-regions where more than one propagating reflected partial wave is allowed ( @xmath98 ) , yielding @xmath99 here , @xmath100 is the unique positive solution of @xmath101 for a fixed @xmath97 and @xmath40 .
the integrand without the exponential factor is related to the fourier transform of @xmath102 .
we plot the three factors determining @xmath102 the magnitudes of the reflection coefficient @xmath103 , the spinor overlap @xmath104 , and the inverse of the parallel - to - defect group - velocity component @xmath105 , as well as their product , in fig .
[ fig : results]d , e , f , and g , respectively . while figs .
[ fig : results]d and e show a featureless dependence on @xmath97 , fig .
[ fig : results]f reveals a peak in @xmath106 . the corresponding local maximum point , which we denote with @xmath107 , is very close to @xmath83 obtained from the numerical result in fig .
[ fig : results]a . as a consequence of the peak in @xmath106 ,
a peak arises in the product of the three factors ( fig .
[ fig : results]g ) as well .
this analysis reveals that the characteristic wave number @xmath83 of the ldos oscillation @xmath73 is determined , to a large extent , by the electronic dispersion relation via @xmath108 , and the details of the scattering process have little significance on its value .
we have repeated the above analysis for various energy values in the range @xmath109 $ ] in order to compare the characteristic wave number @xmath110 of @xmath73 with experimental data @xcite , and to confirm the correlation between the characteristic wave numbers obtained from the dispersion relation [ @xmath111 and from the numerical ldos calculation [ @xmath112 .
we plot @xmath92 as the function of energy @xmath40 in fig .
[ fig : results]c . for low energy @xmath113 mev ,
the hexagonal warping of the cec is weak , and our result shows @xmath114 and a decay of @xmath115 ( not shown in the figures ) , in agreement with the results of the asymptotic analysis @xcite . between 190 mev and 345 mev above the dirac point , our @xmath83 data in fig .
[ fig : results]c differs significantly from @xmath116 , and the former shows good quantitative agreement with the experimental values ( shown in fig .
4b of ref . ) .
remarkably , @xmath117 , shown as green diagonal crosses in fig .
[ fig : results]c , is almost perfectly correlated with @xmath92 , confirming the generality of the above proposition that the characteristic wave number of the ldos oscillation is determined by the electronic dispersion .
no experimental data is available below 190 mev , whereas above 345 mev , the measured data ( shown in fig .
4b of ref . )
shows a pronounced kink that is not described by our model .
a potential reason for that discrepancy is that the surface and bulk conduction electrons might be strongly hybridized in that high - energy range , making our surface - band model inappropriate to describe the corresponding standing - wave patterns . at high energy @xmath118 mev , the nesting of cec segments connected by @xmath95 in fig .
[ fig : results]b may also induce ldos oscillations with wave number @xmath95@xcite . however , in our model we find that such oscillations do not exist , due to the exact cancellation of contributions from reflected and transmitted waves incident from the @xmath44 and @xmath60 regions , respectively ( see sec .
[ sec : cancellation ] ) .
the appearance of the ldos contribution with wave number corresponding to the local maximum point of @xmath106 is a generic feature , expected to be present in other electronic systems as well .
we think that it plays a dominant role in bi@xmath0te@xmath1because of the suppression of the other two fourier components with wave numbers @xmath95 and @xmath119 , due to the cancellation mechanism ( see sec .
[ sec : cancellation ] ) and the absence of backscattering , respectively .
in principle , plane waves incident from the @xmath60 region can contribute to the ldos oscillations in the @xmath44 region , provided that they are transmitted into at least two propagating channels on the @xmath44 side of the line defect . in this section , we show that such a contribution is precisely balanced and canceled out in our model by the interfering reflected components of plane waves incident from the @xmath44 side . in turn , this cancellation is responsible for the absence of ldos oscillations with wave number @xmath95 . without the above cancellation mechanism ,
such oscillations would be expected to arise as @xmath95 connects nesting segments of the cec ( see fig .
as we show now , our above statements follow from the unitary character of the scattering matrix describing the line defect . following the notation used in sec .
[ sec : model ] , consider electron plane waves with energy @xmath40 and parallel - to - defect wave vector component @xmath42 .
we assume for concreteness that for these given parameters @xmath40 and @xmath42 , there exists two incoming , and correspondingly , two outgoing plane waves on the @xmath44 side , and one incoming and one outgoing wave on the @xmath60 side .
the corresponding perpendicular - to - defect wave - vector components are @xmath120 , @xmath121 , @xmath122 , @xmath123 on the @xmath44 side and @xmath124 and @xmath125 on the @xmath60 side , respectively . in this example
, the scattering matrix @xmath126 has the following structure@xcite : @xmath127 a specific example is shown in fig .
[ fig : setup]b , where the points @xmath128 and @xmath129 represent the incoming waves from the @xmath44 region and @xmath130 and @xmath131 represent the reflected and transmitted waves in the @xmath44 region .
note that the conclusions of this section hold for different number of scattering channels as well .
each of the two electron waves incoming from the @xmath44 side is reflected into two propagating states with reflection amplitudes @xmath132 ( @xmath133 ) .
the incident wave from @xmath60 is transmitted into two propagating states on the @xmath44 side with transmission amplitudes @xmath134 ( @xmath135 ) .
straightforward calculation shows that the contribution of the transmitted waves to the ldos oscillations in the @xmath44 region is accompanied by a contribution from the interference of the reflected waves , and the sum of these contributions is proportional to @xmath136 the first factor in eq .
is the scalar product of the first and second rows of the scattering matrix describing the line defect , hence it vanishes because of the unitary character of the scattering matrix .
this finding has the following remarkable consequence with respect to our calculated ldos oscillations @xmath73 . in the @xmath42 intervals where two incoming and two outgoing waves exist ( an example with a specific @xmath42
is shown in fig .
[ fig : setup]b , @xmath128 , @xmath129 , @xmath130 and @xmath131 ) the wave number @xmath137 approaches @xmath95 in a stationary fashion as @xmath42 approaches its extremal value on the cec , which , in principle , implies that the wave number @xmath95 is visible in the ldos oscillations . in practice , however , the prefactor of the term oscillating with @xmath137 , i.e. , the first factor in eq .
, is zero for the whole @xmath42 range with multiple outgoing waves .
note that the wave numbers @xmath138 ( @xmath139 ) , which appear in the ldos oscillations due to interference between incoming and reflected waves , do approach @xmath95 as @xmath42 approaches its extremal value on the cec , but not in a stationary fashion .
consequently , these interference terms are also not able to promote @xmath95 to the dominant wave number of the ldos oscillations . in summary ,
both the theoretical findings presented in this section and our numerical results shown above indicate that in the considered parameter range , the characteristic wave number of the ldos oscillation in the vicinity of the line defect is not @xmath95 .
a line defect on a metallic surface is usually modeled as a potential step @xcite or as a localized potential barrier @xcite ( e.g. , dirac - delta potential ) . in our work
, we make the convenient , but arbitrary choice of modeling the defect as a potential step .
it is important to note that the fine details of the ldos results might in fact depend on the choice of the model of the defect ( step vs. localized barrier ) .
however , our interpretation explaining our main result , i.e. , the dependence of the standing wave s wave number @xmath140 on the fermi energy @xmath40 , is based on the momentum - dependence of the inverse velocity @xmath141 ( see figs . [
fig : results]c - g ) .
the latter quantity is independent of the model describing the line defect , therefore our main conclusion is expected to hold even if ( a ) the height of the potential step at the line defect , denoted by @xmath142 , is varied , or ( b ) a different model ( e.g. , dirac - delta potential barrier ) is used to represent the defect . to further support the expectation ( a )
, we have numerically calculated the ldos oscillations for various values of the potential step height @xmath142 and found no qualitative change in the inferred @xmath140 data .
mev for the case when the line defect is perpendicular to the @xmath6 direction , and the fit of @xmath143 with fitting parameters @xmath144 , @xmath97 and @xmath145 ( solid red line ) .
( b ) constant energy contour and the relevant wave numbers @xmath95 and @xmath146 .
the wave number of the oscillation in ( a ) is given by @xmath147 , as the oscillations with @xmath146 are fast decaying ( @xmath148 ) due to the absence of backscattering.,scaledwidth=48.0% ] even though the line defect in the considered experiment ref .
was perpendicular to the @xmath5 direction of the surface brillouin zone@xcite , it is instructive , and regarding future experiments , potentially useful to consider the other high - symmetry case when the line defect is oriented perpendicular to the @xmath6 direction . in fig .
[ fig : ldos_gk ] , we demonstrate that both the wave vector and the decay characteristics of the ldos standing waves we obtain from our numerical technique are in complete correspondence with the analytical results of asymptotic analysis@xcite .
namely , the wave number of the oscillation is given by the extremal ( maximal ) perpendicular - to - defect width @xmath147 of the constant energy contour , whereas the decay goes as @xmath149 .
oscillations with @xmath146 are not seen in fig .
[ fig : ldos_gk]a presumably because they are fast decaying ( @xmath148 ) due to the absence of backscattering@xcite . in conclusion , we theoretically described pre - asymptotic electronic ldos oscillations in the vicinity of a line defect on the surface of bi@xmath0te@xmath1 , with wave number and decay characteristics markedly different from the asymptotic ones .
the calculated energy dependence of the characteristic wave number of the ldos oscillations is in line with stm data . in a general context
, our study highlights the importance of pre - asymptotic calculation of the surface - state ldos oscillations in the analysis and interpretation of stm experiments .
_ note : _ while completing this manuscript , we became aware of a related work @xcite on electronic standing waves on 3dti surfaces .
our results partially overlap with those in ref . .
apart from various details of the model , the major distinctive features of our work are ( i ) the interpretation of the results in terms of the properties of the group velocity ( fig . [
fig : results]d - g ) , and ( ii ) the quantitative agreement with the experimental results reported in ref . .
this work was supported by otka grants no .
75529 , no . 81492 , and no .
pd100373 , the marie curie itn project nanoctm , the european union and co - financed by the european social fund ( grant no .
tamop 4.2.1/b-09/1/kmr-2010 - 0003 ) , and the marie curie grant cig-293834 .
to base our calculation on an accurate surface - state dispersion relation , in sec .
[ sec : parameters ] .
we estimated the band - structure parameters of bi@xmath0te@xmath1 by comparing the theoretical dispersion ( see eq . ) and dos with the experimentally observed arpes and stm spectra and the dos derived from those .
the four band - structure parameters are @xmath150 , @xmath151 , @xmath31 and @xmath20 . here
we outline the considerations we used for those estimates .
the signs of @xmath31 and @xmath20 can be determined by considering the arpes spectrum along the @xmath5 direction , shown as red crosses in fig .
[ fig : dos]a and b. note that this cut of the dispersion relation corresponds to the function @xmath152 in eq . .
the measured dispersion is linear for small wave number , its slope first becomes smaller as the wave number is increased , but then the slope increases again as wave number is increased further .
this characteristic is naturally captured by a polynomial of the wave number with negative second - order coefficient and positive third - order coefficient .
since the third - order taylor series of @xmath152 in @xmath22 around zero has the form @xmath153 we can conclude that @xmath154 and @xmath155 is required to describe the measured dispersion .
the signs of the remaining two parameters @xmath150 and @xmath151 has no effect on the spectral properties , therefore we assign a positive sign to them . having the signs of band - structure parameters
established , we determined the values of the four parameters by systematic visual comparison of the experimental dispersions ( figs .
[ fig : dos]a , b ) , the dos data obtained from arpes and stm measurements ( figs . 1c , d of ref . ) , and the corresponding theoretical curves .
we label the electronic plane - wave states by their energy @xmath40 and the parallel - to - defect wave - number component @xmath42 , which are conserved in the scattering process on the line defect ( for details see the main text ) . here
we review a numerical method to obtain these states , which is necessary to solve the scattering problem at the edge step . for a given @xmath40 and @xmath42
there are six solutions of longitudinal wave vector @xmath47 ( @xmath156 ) which satisfies the characteristic equation @xmath157 = 0 $ ] , where @xmath158 is the @xmath159 identity matrix and @xmath26 is the hamiltonian defined by eq .
( 1)in the main text .
complex roots @xmath47 of the characteristic polynomial @xmath160 = \sum\limits_{k=0}^6a(e , q)k_{r}^k\;\label{eq : polynom}\ ] ] are equal to eigenvalues of the companion matrix @xcite of this polynomial , hence we find the roots @xmath47 by numerically diagonalizing the companion matrix .
then , the spinor component @xmath161 of the corresponding plane wave @xmath162 can be numerically computed from the eigenvalues problem @xmath163 for @xmath164 .
shown in fig .
[ fig : results]a , ( b ) the function @xmath165 , where @xmath82 , @xmath83 , @xmath84 , @xmath85 are obtained from fitting @xmath166 to @xmath167 , ( c ) @xmath168 , and ( d ) @xmath169 . , scaledwidth=48.0% ] in sec .
[ sec : results ] , we present numerical results for the defect - induced spatial modulation of the ldos , @xmath73 . in order to develop an understanding of the decay characteristic of the ldos oscillations ,
we fit a power - law decaying ( @xmath170 ) as well as an exponentially decaying function , @xmath171 and @xmath166 , respectively , to our data . according to fig .
[ fig : results]a , the exponentially decaying @xmath166 provides a better fit , hinting that the decay characteristics is closer to an exponential than to a power - law predicted earlier@xcite .
however , our fitting procedure is not conclusive , as the exponentially decaying function @xmath166 has an extra fitting parameter , the length scale @xmath85 of the decay . in order to investigate the decay characteristics further ,
here we provide the discrete fourier transform ( ft ) ( fig .
[ fig : fft]a ) of the real - space data in fig .
[ fig : results]a , and compare that to discrete fourier transformed oscillations that decay in an exponential ( fig .
[ fig : fft]b ) or power - law ( fig [ fig : fft]c , d ) fashion .
the discrete fourier transformation is carried out after symmetrization of the real - space data , i.e. , after mapping the real - space data set @xmath172 @xmath173 to @xmath174 ( @xmath175 ) via the definition @xmath176 this symmetrization ensures that the fourier transform will be real valued in the large @xmath177 limit . the subplots of fig .
[ fig : fft ] show the ft of ( a ) our numerical results @xmath93 shown in fig .
[ fig : results]a , ( b ) the function @xmath165 , where @xmath82 , @xmath83 , @xmath84 , @xmath85 are obtained from fitting @xmath166 to @xmath167 , ( c ) @xmath168 , and ( d ) @xmath169 . the data in fig .
[ fig : fft]c , d shows , in accordance with the known analytical formula describing the fourier transform of power - law decaying sinusoidal oscillations@xcite , that the ft develops non - analytical behavior at the characteristic wave number @xmath110 .
in contrast , our data set in fig . [
fig : fft]a shows no such non - analytical behavior , similarly to the ft of the exponentially decaying oscillation in fig . [ fig : fft]b .
this observation , although still not conclusive , further supports the possibility that the ldos oscillations in the vicinity of the line defect do not follow a power - law decay , and perhaps are closer to an exponential . further analytical studies , e.g. ,
the extension of the asymptotic analysis @xcite to the pre - asymptotic spatial region might help settling this open issue . in sec .
[ sec : results ] , we argued that the pre - asymptotic component of the ldos is due to the interference of incoming and reflected partial waves i.e. , that the role of evanescent and transmitted partial waves is negligible . to strengthen that point further
, we plot the quantity @xmath178 as a function of @xmath97 [ for definitions , see around eq . ] , as a dashed blue line in fig .
[ fig : fft]a .
note that @xmath179 is the fourier transform of the symmetrized @xmath102 of eq . .
figure [ fig : fft]a further supports the interpretation that the dip around the characteristic wave number @xmath92 of the ldos oscillation forms as a result of interference of incoming and reflected partial waves .
99 m. z. hasan and c. l. kane , rev .
82 * , 3045 ( 2010 ) .
d. hsieh , d. qian , l. wray , y. xia , y. s. hor , r. j. cava and m. z. hasan , nature * 452 * , 970 ( 2008 ) .
d. hsieh _ et al .
_ , science * 323 * , 919 ( 2009 ) . y. xia et al . ,
nat . phys . * 5 * , 398 ( 2009 ) .
y. l. chen _
et al . _ , science * 325 * , 178 ( 2009 ) .
d. hsieh et al . , phys .
lett . * 103 * , 146401 ( 2009 ) . k. kuroda , m. arita , k. miyamoto , m. ye , j. jiang , a. kimura , e. e. krasovskii , e. v. chulkov , h. iwasawa , t. okuda , k. shimada , y. ueda , h. namatame and m. taniguchi phys . rev
. lett . * 105 * , 076802 ( 2010 ) .
p. roushan , j. seo , c. v. parker , y. s. hor , d. hsieh , d. qian , a. richardella , m. z. hasan , r. j. cava and a. yazdani , nature * 460 * , 1106 ( 2009 ) .
et al . _ ,
103 * , 266803 ( 2009 ) . j. seo , p. roushan , h. beidenkopf , y. s. hor , r. j. cava and a. yazdani , nature * 466 * , 343 ( 2010 ) .
z. alpichshev , j. g. analytis , j .- h . chu , i. r. fisher , y. l. chen , z. x. shen , a. fang and a. kapitulnik , phys .
rev . lett . * 104 * , 016401 ( 2010 ) .
j. wang , w. li , p. cheng , c. song , t. zhang , p. deng , x. chen , x. ma , k. he , j .- f .
jia , q .- k .
xue , b .- f .
zhu , phys .
b * 84 * , 235447 ( 2011 ) .
m. f. crommie , c. p. lutz and d. m. eigler , nature * 363 * , 524 ( 1993 ) .
l. fu , phys .
* 103 * , 266801 ( 2009 ) .
lee , c. wu , d. p. arovas and s .- c .
zhang , phys .
b * 80 * , 245439 ( 2009 ) . h. m. guo and m. franz , phys . rev .
b * 81 * , 041102(r ) ( 2010 ) . q
wang , d. wang and f .- c .
zhang , phys .
b * 81 * , 035104 ( 2010 ) .
x. zhou , c. fang , w. f. tsai , and j. p.
hu , phys .
b * 80 * , 245317 ( 2009 ) . r. r. biswas and a. v. balatsky , phys . rev .
b * 83 * , 075439 ( 2011 ) . r. r. biswas and a. v. balatsky , arxiv:1005.4780 ( unpublished ) . q. liu , arxiv:1108.6051 ( unpublished ) . s. basak , h. lin , l. a. wray , s .- y .
xu , l. fu , m. z. hasan , and a. bansil , phys .
b * 84 * , 121401(r ) ( 2011 ) .
s. datta , _ electronic transport in mesoscopic systems _ ( cambridge university press , 1995 ) .
d. zhang and c. s. ting , phys .
b * 85 * , 115434 ( 2012 ) .
r. a. horn and c. r. johnson , _ matrix analysis _
( cambridge university press , 1985 ) . | a line defect on a metallic surface induces standing waves in the electronic local density of states ( ldos ) .
asymptotically far from the defect , the wave number of the ldos oscillations at the fermi energy is usually equal to the distance between nesting segments of the fermi contour , and the envelope of the ldos oscillations shows a power - law decay as moving away from the line defect . here
, we theoretically analyze the ldos oscillations close to a line defect on the surface of the topological insulator bi@xmath0te@xmath1 , and identify an important preasymptotic contribution with wave - number and decay characteristics markedly different from the asymptotic contributions .
the calculated energy dependence of the wave number of the preasymptotic ldos oscillations is in quantitative agreement with the result of a recent scanning tunneling microscopy experiment [ phys .
rev .
lett . * 104 * , 016401 ( 2010 ) ] . |
the `` teleparallel '' equivalent of general relativity ( tegr ) @xcite is an equivalent formulation of classical gravity , in which instead of using the torsionless levi - civita connection one uses the curvatureless weitzenbck one .
the dynamical objects are the four linearly independent vierbeins ( these are _ parallel _ vector fields represented by the appellation `` teleparallel '' ) .
the advantage of this framework is that the torsion tensor is formed solely from products of first derivatives of the tetrad .
as described in @xcite , the lagrangian density @xmath0 can be constructed from this torsion tensor under the assumptions of invariance under general coordinate transformations , global lorentz transformations , and the parity operation , along with requiring the lagrangian density to be second order in the torsion tensor .
thus , apart from possible conceptual differences , tegr is completely equivalent and indistiguishable form general relativity ( gr ) at the level of equations , both background and perturbation ones .
on the other hand , in general relativity one can add the quintessence scalar field in order to acquire a dynamical dark energy sector , a scenario that exhibits a very interesting cosmological behavior and has gained a huge amount of research @xcite . amongst others
, one can generalize it by including a nonminimal coupling between the quintessence field and gravity @xcite , or more generally extend it to the scalar - tensor paradigm @xcite .
one can also use a phantom instead of a canonical field @xcite , or the combination of both these fields in a unified scenario called quintom @xcite . in this letter
we are interested in formulating `` teleparallel dark energy '' , adding a canonical scalar field in tegr . in the minimal coupling case
the resulting theory is identical to the ordinary quintessence , both at the background and perturbation levels .
however , when the nonminimal coupling is switched on , teleparallel dark energy is different from its gr counterpart , and the cosmological behavior of such a new scenario proves to be very interesting .
let us review very briefly the quintessence paradigm in general relativity .
in such a scenario the dark energy sector is attributed to a homogeneous scalar field @xmath1 , and the action is given by @xcite @xmath2 \label{action},\ ] ] with @xmath3 , @xmath4 , @xmath5 the scalar - field potential , @xmath6 the non - minimal coupling parameter , @xmath7 the ricci scalar , and @xmath8 the matter lagrangian . note the difference in the metric signature that exists amongst the various works in the literature and the corresponding sign changes in the action , since a change in the metric signature leads @xmath9 , @xmath10 and @xmath11 to change sign , while @xmath7 and the energy - momentum tensor remain unaffected @xcite . in this letter
we use the signature @xmath12 in all sections , just to be closer to the literature of teleparallel gravity .
thus , under this convention , the conformal value of @xmath6 is @xmath13 . in the case of a flat friedmann - robertson - walker ( frw ) background metric @xmath14 where @xmath15 is the cosmic time , @xmath16 are the comoving spatial coordinates and
@xmath17 is the scale factor , the friedmann equations write as @xmath18 where @xmath19 is the hubble parameter , a dot denotes differentiation with respect to @xmath15 , and @xmath20 and @xmath21 are the matter energy density and pressure , respectively , following the standard evolution equation @xmath22 , with @xmath23 the matter equation - of - state parameter .
additionally , we have introduced the energy density and pressure of the nonminimally coupled scalar field , given by @xcite : @xmath24 where a prime denotes derivative with respect to @xmath1 .
we note that the above relations have been simplified by using the useful expression @xmath25 in the frw geometry .
as mentioned above , in such a scenario , dark energy is attributed to the scalar field , and thus its equation - of - state parameter reads : @xmath26 finally , the equations close by considering the evolution equation for the scalar field @xcite : @xmath27 which can alternatively be written in the standard form @xmath28 .
we now briefly review tegr .
the notation is as follows : greek indices @xmath29 ... and capital latin indices @xmath30 ... run over all coordinate and tangent space - time 0 , 1 , 2 , 3 , while lower case latin indices ( from the middle of the alphabet ) @xmath31 and lower case latin indices ( from the beginning of the alphabet ) @xmath32 ... run over spatial and tangent space coordinates 1 , 2 , 3 , respectively . as stated in _
introduction _ , the dynamical variable of `` teleparallel '' gravity is the vierbein field @xmath33 .
this forms an orthonormal basis for the tangent space at each point @xmath34 of the manifold , that is @xmath35 , where @xmath36 .
furthermore , the vector @xmath37 can be analyzed with the use of its components @xmath38 in a coordinate basis , that is @xmath39 .
in such a construction , the metric tensor is obtained from the dual vierbein as @xmath40 contrary to gr , which uses the torsionless levi - civita connection , in tegr ones takes the curvatureless weitzenbck connection @xcite , whose torsion tensor reads @xmath41 where @xmath42 .
moreover , the contorsion tensor , which equals to the difference between weitzenbck and levi - civita connections , is defined as @xmath43 and we also define @xmath44 . in the present formalism
all the information concerning the gravitational field is included in the torsion tensor @xmath45 . using the above quantities one can extract the form of the `` teleparallel lagrangian '' , which is nothing else than the torsion scalar , namely @xcite : in summary , the simplest action in a universe governed by teleparalel gravity is @xmath46,\end{aligned}\ ] ] where @xmath47 ( one could also include a cosmological constant ) . variation with respect to the vierbein fields gives equation of motion @xmath48 where @xmath49 stands for the usual energy - momentum tensor .
these equations are exactly the same as those of gr for every geometry choice .
in particular , for the frw background metric ( [ frwmetric ] ) , the vierbein choice of the form @xmath50 is an exact solution @xcite of the field equation in eq .
( [ eom ] ) , which does not generate a divergent energy for the whole space - time .
furthermore , it is easily seen that the corresponding friedmann equations are identical to the gr ones , both at the background and perturbation levels @xcite .
let us now construct teleparalell dark energy . this will be done by adding a scalar field in the equivalent , teleparallel , formulation of gr .
thus , the action will simply read : @xmath51 .
\label{action2}\ ] ] we emphasize that in the above action a nonminimal coupling between the scalar field and gravity is allowed .
although in the nonminimal case one could use alternative torsion scalars , we prefer to keep the standard one for simplicity .
we also note that the action in ( [ action2 ] ) with the torsion formulation of gr is similar to the standard nonminimal quintessence where the scalar field couples to the ricci scalar . in this scenario ,
similar to the standard quintessence , dark energy is attributed to the scalar field , and thus its equation - of - state parameter ( @xmath56 ) is defined to be the same as that in ( [ eos ] ) , but @xmath57 and @xmath58 are now given by ( [ telerho ] ) and ( [ telep ] ) , respectively .
finally , one can see that the scalar field evolution ( [ fieldevol2 ] ) leads to the standard relation @xmath59 .
we now explore the cosmological implications of the scenario at hand .
firstly , we immediately observe that in the case of the minimal coupling , teleparallel dark energy coincides with quintessence ( see ( [ rhophi])-([pphi ] ) and ( [ telerho])-([telep ] ) ) , and one can verify that at the level of perturbations too .
this is expected since , concerning the gravitational sector , tegr is identical with gr , and in the minimal case one just adds a distinct scalar sector , thus making no difference whether it is added in either of the two theories .
however , things are different if we switch on the nonminimal coupling . in this case the additional scalar sector is coupled to gravity , with the curvature scalar in gr and with the torsion scalar in tegr , and thus the resulting coupled equations do not coincide .
clearly , teleparallel dark energy , under the nonminimal coupling , is a different theory .
let us proceed in presenting some basic and general features of the nonminimal coupling of the scalar - torsion theory .
apart from the straightforward results that dark energy possesses a dynamical nature as well as it can drive the universe acceleration , the most interesting and direct consequence of the dark energy density and pressure relations ( [ telerho])-([telep ] ) is that the dark energy equation - of - state parameter can lie in the quintessence regime ( @xmath60 ) , in the phantom regime ( @xmath61 ) , or exhibit the phantom - divide crossing during cosmological evolution .
this is a radical difference with the quintessence scenario and reveals the capabilities of the construction . in order to present the above features in a more transparent way
, we evolve numerically the cosmological system for dust matter ( @xmath62 ) , using the redshift @xmath63 as the independent variable , imposing the present scale factor @xmath64 to be equal to 1 , the dark energy density @xmath65 at present to be @xmath66 and its initial value to be @xmath67 . finally , concerning the scalar field potential we use the exponential ansatz of the form @xmath68 . in fig
[ plot ] we depict the @xmath56-evolution for three realizations of the scenario at hand . in the case of the black - solid curve
the teleparallel dark energy behaves like quintessence , in the red - dashed curve it behaves like a phantom , while in blue - dotted curve the dark energy exhibits the phantom - divide crossing during the evolution .
note that the crossing behavior in fig .
[ plot ] is the one favored by the observational data , in contrast with viable @xmath69-gravity models where it is the opposite one @xcite .
we remark that in the above graphs we focus on their qualitative features , and in particular we maintain the same potential just to stress that in principle one can obtain the various behaviors with the same potential .
clearly , one could be quantitatively more accurate and impose the observational @xmath70 as an input , reconstructing the corresponding potential .
however in the present work we desire to remain as general as possible .
in the present scenario of `` teleparallel '' dark energy we have added a scalar field to the teleparallel equivalent to general relativity ( tegr ) , allowing for a nonminimal coupling between the field and gravity . in the minimally - coupled case
the cosmological equations coincide with those of the standard quintessence .
however when the nonminimal coupling is switched on the resulting theory exhibits different behavior .
in particular , although the scalar field is canonical , one can obtain a dark energy sector being quintessence - like , phantom - like , or experiencing the phantom - divide crossing during evolution , a behavior that is much richer comparing to general relativity ( gr ) with a scalar field .
moreover , the fact that the phantom regime can be described without the need of phantom fields , which have ambiguous quantum behavior @xcite , is a significant advantage .
the physical reason for the aforementioned difference , despite the equivalence of pure gr and pure tegr , is that while in gr one couples the scalar field with the only suitable gravitational scalar , namely the ricci scalar @xmath7 , in the later one couples the scalar field with the only suitable gravitational scalar , namely the torsion scalar @xmath0 .
the richness of the resulting theory comparing to gr quintessence is additionally manifested in the fact that , although in the later one can perform a conformal transformation and transit to an `` equivalent '' , minimally - coupled , theory with transformed field and potential @xcite , in the former such a transformation does not exist since one obtains extra terms depending on the torsion tensor itself , as can be easily verified transforming the vierbeins as @xmath71 ( one applies in our case the similar analysis of @xcite of the case of @xmath72 scenarios ) . thus , teleparallel dark energy can not be transformed to an `` equivalent '' minimally coupled form , which is known to be able to describe only the quintessence regime , and this indicates its richer structure .
such an absence of conformal transformation exists in other cosmological scenarios too , for example in scalar - field models with non - minimal derivative couplings , where it is also known that the resulting theories possess a richer structure @xcite .
the addition of a scalar field to tegr was inspired by the corresponding procedure in gr .
however , although in gr one can alternatively and equivalently generalize the action to @xmath69 , freeing himself of the need to add the scalar field , in the teleparallel formulation of gr the generalization to @xmath72 @xcite seems to spoil the local lorentz invariance for all functions apart from the linear one @xcite .
however , at the background level no new degrees of freedom are present , while at linear perturbation the new vector degree of freedom only satisfies constraint equations @xcite .
similarly , in our generalization of tegr , in the case of non - minimal coupling , a lorentz - violating term appears ( the last term in the left hand side of ( [ eom2 ] ) ) , despite the fact that the theory is linear in @xmath0 .
however , no new degree of freedom will appear at the background level on which we focus on this work . clearly , going beyond background evolution and examine whether the lorenz violations do indeed appear under cosmological geometries and scales ( we have checked that at the low - energy limit , the theory s basic parametrized post newtonian parameters are consistent with solar system observations ) , and if they can be detected , is an interesting and open subject , as it is in @xmath72 gravity too , and
will be incorporated in more details elsewhere . in summary ,
the rich behavior of teleparallel dark energy makes it a promising cosmological scenario . in this work
we have desired to remain as general as possible , and present its basic and novel features . clearly , before it can be considered as a good candidate for the description of nature
, one needs to investigate various subjects , such as to perform a detailed perturbation analysis , to use observational data in order to constrain the parameters of the model , to examine the phase - space behavior in order to reveal the late - time cosmological features , etc .
such aspects , although necessary , lie outside the goal of the present work and are left for future investigations .
a. einstein 1928 , sitz .
p. 217 ; ibid
p. 224 ; a. unzicker and t. case , arxiv : physics/0503046 . k. hayashi and t. shirafuji , phys .
d * 19 * , 3524 ( 1979 ) ; addendum - ibid .
* 24 * , 3312 ( 1982 ) .
v. sahni and s. habib , phys .
lett . * 81 * , 1766 ( 1998 ) ; j. p.
uzan , phys . rev .
d * 59 * , 123510 ( 1999 ) ; n. bartolo , m. pietroni , phys .
rev . * d61 * , 023518 ( 2000 ) ; v. faraoni , phys .
d * 62 * , 023504 ( 2000 ) ; e. elizalde , s. nojiri and s. odintsov , phys . rev .
d * 70 * , 043539 ( 2004 ) .
j. w. maluf , j. math .
* 35 * ( 1994 ) 335 ; h. i. arcos and j. g. pereira , int .
j. mod .
d * 13 * , 2193 ( 2004 ) .
k. bamba , c. q. geng and c. c. lee , jcap * 1011 * , 001 ( 2010 ) . j. m. cline , s. jeon and g. d. moore , phys . rev .
d * 70 * , 043543 ( 2004 ) .
l. amendola , phys .
b * 301 * , 175 ( 1993 ) ; s. f. daniel and r. r. caldwell , class .
* 24 * , 5573 ( 2007 ) ; e. n. saridakis , s. v. sushkov , phys .
rev . * d81 * , 083510 ( 2010 ) ; c. germani and a. kehagias , phys .
lett . * 105 * , 011302 ( 2010 ) .
g. r. bengochea and r. ferraro , phys .
d * 79 * , 124019 ( 2009 ) ; e. v. linder , phys .
d * 81 * , 127301 ( 2010 ) ; k. bamba , c. q. geng , c. c. lee and l. w. luo , jcap * 1101 * , 021 ( 2011 ) ; s. -h . chen , j. b. dent , s. dutta , e. n. saridakis , phys . rev . * d83 * , 023508 ( 2011 ) ; k. bamba , c. q. geng and c. c. lee , arxiv:1008.4036 [ astro-ph.co ] .
b. li , t. p. sotiriou , j. d. barrow , phys .
rev . * d83 * , 064035 ( 2011 ) .
b. li , t. p. sotiriou , j. d. barrow , phys .
* d83 * , 104017 ( 2011 ) ; m. li , r. x. miao and y. g. miao , arxiv:1105.5934 [ hep - th ] . | using the `` teleparallel '' equivalent of general relativity as the gravitational sector , which is based on torsion instead of curvature , we add a canonical scalar field , allowing for a nonminimal coupling with gravity . although the minimal case is completely equivalent to standard quintessence , the nonminimal scenario has a richer structure , exhibiting quintessence - like or phantom - like behavior , or experiencing the phantom - divide crossing .
the richer structure is manifested in the absence of a conformal transformation to an equivalent minimally - coupled model . |
the be binary system ( associated with the radio source gt 0236 + 610 ) is remarkable for dramatic radio outbursts occurring with the 26.5 day orbital cycle ( @xcite ) , and for its possible association with the 100 mev gamma - ray source 2cg 135 + 01 ( @xcite ) .
radio flares lasting several days occur every orbit , with the peak flux varying in phase by up to half the orbital cycle .
the binary is a weak , variable x - ray source ( @xcite ) with a non - thermal spectrum .
egret measurements indicate that the 100 mev emission from the region may also be variable ( @xcite ) .
the gamma - ray error box contains no other likely counterparts .
the gamma - ray localization is not , however , sufficiently accurate to make a secure association between 2cg 135 + 01 and . the mechanisms responsible for the radio outbursts and variable x - ray emission are not well understood .
the radio emission is consistent with optically - thin synchrotron radiation for most of the outburst , with indication that the source is self - absorbed at the beginning of the outburst rise ( @xcite ) .
the x - ray emission is consistent with either an extension of the radio synchrotron spectrum , or inverse compton scattering of the stellar photons by the relativistic electrons responsible for the radio flux ( @xcite ) .
the most plausible models proposed for the origin of the relativistic electrons are acceleration in a shock produced at the interaction of the be star wind with the relativistic wind of a young pulsar companion ( @xcite ) , or in a shock produced by the ejection of matter accreted onto a pulsar magnetosphere ( @xcite ) . in either case , both the x - ray and radio emission would vary with orbital phase . in this paper , we present simultaneous _ rxte _ x - ray , green bank interferometer ( gbi ) and ryle telescope radio observations taken over a single orbital cycle .
these observations provide a complete orbital light curve in 210 kev x - rays and 2.25 ghz , 8.3 ghz , and 15 ghz radio frequencies . on only one prior occasion
has been monitored in the x - ray band over an entire orbital cycle ( @xcite ) .
we have also analyzed archival x - ray data and compiled a history of the x - ray variability .
we have established x - ray flux variations on hourly timescales , and have performed the most sensitive searches to date for rapid x - ray pulsations from the suggested neutron star companion .
in addition , we have analyzed the 2 150 kev spectra of both and the nearby quasar qso 0241 + 622 , which has been included in the field of view of many previous hard x - ray measurements .
.observation log of rxte observations of , with fitted fluxes . [ cols="^,^,^,^,^,^,^,^ " , ] in addition to the _ rxte _ search , we performed a pulsation search on the _ asca _ data .
for the gis detectors during p1 the time resolution of 0.0625 s limits the search to a minimum period of 0.125 s , and for p2 , the minimum period searched is 1.0 s , corresponding to the nyquist limit for the 0.5 s sampling .
epoch folding using the standard xronos software ( available from the high energy astrophysics science archive research center ( heasarc ) at goddard space flight center ) failed to detect periodicity in either pointing .
table 2 summarizes the 90% confidence upper limits on the amplitude of pulsed emission for searches on the separate and combined gis 1 and 2 data .
we calculated these limits using the formalism described in leahy _
et al . _ ( 1983 ) . a second source in the asca fov , identified as a stellar source ( labeled source2 ) was also searched for pulsations , and the upper limits are also given in table 2 .
for p1 , in addition to the epoch folding , we performed an fft , which yielded nearly identical upper limits on the pulsed amplitude fraction as the epoch folding . for p2 ,
aperiodic variability in the source precluded the use of an fft for period searching ( see [ sec - aperiodic ] ) .
we also searched the sis data to a minimum period of 8 s , ( determined by the 4 s ccd readout time ) , and found no pulsations , with limits similar to those presented for the gis . to search for short - period pulsations ,
for which smearing of the pulsed signal due to variable doppler shifts resulting from the binary orbital motion can be important , we performed an acceleration search on the gis data from p1 .
the variation in the pulse frequency is well - approximated by a linear trend over the duration of each observation .
we searched a range of constant frequency derivatives from @xmath1s@xmath2 ( consistent with the range expected from the orbital solutions ) in 1600 spacings .
no periodicity was detected .
several common characteristics of the two simultaneous , well - sampled x - ray and radio lightcurves ( figures [ fig - xtelight ] and [ fig - rosatlight ] ) are evident .
both are clearly modulated at the orbital period , and the peak in the x - ray emission is significantly offset ( by 0.4 0.5 in phase ) from the peak in the radio emission .
although containing only two data points , the _ asca _ observations are also consistent with the anticorrelation of radio and x - ray emission . in the _ rosat _ 0.5 2 kev data ,
the ratio of peak to quiescent luminosity is a factor @xmath320 , similar to what is observed in the radio , while in the _ rxte / pca _ 210 kev data , the variation is a factor 3 7 ( the range depending on the uncertain zero point ) . because of the poor spectral resolution and narrow bandpass of the _ rosat _ data , and the difficulty in determining spectral parameters in the _ rxte / pca _ data , it is impossible to determine if this difference arises from variable x - ray absorption ( potentially important in the soft _ rosat _ band but negligible for the pca ) , variation in the spectral index , or variation in the x - ray outburst intensity orbit - to - orbit .
these observations suggest that it is unlikely that both the x - ray and the radio flux arise from synchrotron emission from a common electron population .
rather , it is more likely that the x - ray flux results from inverse compton scattering of stellar photons off a quasi - steady relativistic electron population .
first , it is difficult to accomodate a significant offset between the x - ray and radio peaks if both result from synchrotron emission from a common population .
furthermore , for reasonable synchrotron models for the radio emission , inverse compton losses should dominate at x - ray wavelengths over the entire orbit .
for example , peracaula ( 1997 ) has modeled the synchrotron emission from an expanding spherical plasmon ejected from the binary system to explain the radio emission . to fit these data ,
the initial radius is @xmath4 cm , the distance from the star is the orbital separation ( @xmath5 cm ) , the magnetic field is @xmath6 g , and the expansion velocity is @xmath7 cm / s .
electrons are injected uniformly into the expanding sphere at a steady rate over 7 days . in this model , to get the radio rise requires a large initial @xmath8 , and for the implied magnetic field and @xmath8 , inverse compton losses will dominate over synchrotron losses at all distances from the star . for synchrotron losses to be important at x - ray energies
requires @xmath9 g for an electron population with spectral index 2.1 ( steeper spectral index requires a higher field ) .
the small - amplitude of the orbital modulation of the x - ray flux relative to the radio observed by _
rxte _ indicates that the electron population responsible for the inverse compton emission is relatively constant , with the flux modulation likely resulting from variations in the stellar photon density as a function of position in an eccentric orbit .
the x - ray maximum will occur near periastron , where the stellar photon density is highest ( although some offset is possible if the be star equator is not aligned with the orbit ) .
the factor 3 6 x - ray variations would imply @xmath10 , or orbital eccentricity @xmath11 for a steady electron population .
this range is consistent with the poorly - defined orbital parameters , which allow a range of eccentricity from 0.2 0.8 ( @xcite ) .
as originally proposed by marashi and treves ( 1981 ) , a relatively steady electron population could be produced at the shock resulting from the interaction of the relativistic wind of a young pulsar with the be stellar wind .
the pulsar wind would create a cavity , with the boundary determined by dynamic pressure balance .
these electrons inverse compton scatter the stellar photons , producing x - ray emission , modulated by the variation in photon flux due to an eccentric orbit .
the observed 2 10 kev x - ray luminosity of @xmath12 erg s@xmath13 is consistent with the shock interpretation , and is similar to the observed luminosity near periastron of psr b1259@xmath1463 ( @xcite ) , for which a similar mechanism has been invoked ( @xcite ) .
this model is also consistent with the lack of observed x - ray pulsations , since the x - ray emission is produced well outside the pulsar magnetosphere . to explain the periodic radio outbursts
requires an expanding plasmon .
this could be produced when , sometime after periastron , the stellar wind pressure at the back side of the cavity becomes low enough that the cavity is no longer closed , allowing relativistic particles to stream away from the system , producing the radio emission and the observed expanding vlbi source . in this scenario ,
the x - ray peak should occur at periastron , and the radio outburst approximately 0.5 in phase away , as is observed .
if 2cg 135 + 01 is associated with the system , the 100 mev gamma - rays could also be produced via the inverse compton mechanism by electrons of energy @xmath15 erg .
this emission should also be modulated at the orbital period .
the egret flux from 2cg 135 + 01 is , in fact , time variable on day timescales ( @xcite ) , however the statistics are not sufficient to establish an orbital modulation .
numerous instruments have claimed detection of at hard x - ray and gamma - ray energies , however few have had positioning capability sufficient to distinguish between emission from the binary and qso 0241 + 622 ( qso 0241 + 622 being the only source within a several degree radius with a known hard spectrum ) .
figure [ fig - hespec ] summarizes some of the most secure hard x - ray and gamma - ray detections from this region . as described previously ,
the pca and hexte clearly detect both sources . in the pca ,
the quasar 2 10 kev flux is a factor @xmath32 higher than the average high - state of . in hexte , the quasar 15 40 kev flux ( determined at a different epoch than the pca quasar measurement ) is a factor 2 7 higher than the hexte detection .
this would tend to indicate variability of the qso , however the statistics are limited . to derive pca fluxes for , given the uncertainties in the pca background subtraction and response for the pointing
, we fit a power law model with photon index 1.8 ( consistent with the _ asca _ values ) , and we show a range that includes uncertainty in the zero level . for the hexte quasar detection , we used the standard background subtraction , and fit a flat spectrum over the 15 - 40 kev band to determine the flux .
the osse and comptel measurements shown in the figure included both and qso 0241 + 622 in the instrument fields of view ( @xcite).the egret angular resolution is sufficient to exclude the quasar as the source of the gamma - ray emission ( but the 13 error radius makes positive association of the 100 mev flux with uncertain ) .
it is clear from the _ rxte _ observations that both the osse and comptel measurements were dominated by flux from the quasar .
both observations integrated over times comparable to the orbit , and should be consistent with the average flux , which is a factor @xmath16 lower than the qso flux .
our observations have confirmed that the x - ray and soft gamma - ray emission from is variable , and modulated on the orbital timescale .
the _ rxte _
observations sample a harder band than the previous _ rosat _ data , and we can eliminate variable absorption as the source of the flux modulation . from the two well - sampled lightcurves ,
it is clear that the x - ray emission peaks almost half an orbit before the radio .
one explanation for this offset is that the high - energy emission is produced by inverse compton scattering of stellar photons by a population of relativistic electrons produced at a nearly steady rate at the shock interaction of the relativistic wind of a young pulsar and the be star wind .
this would produce an x - ray peak near periastron .
the synchrotron radio emission would then arise from the expansion of the plasma near apastron , when the cavity produced by the pulsar is no longer confined .
our observations fail to detect x - ray pulsations that would signify the presence of a young neutron star .
we also detected x - ray emission from the nearby qso 0241 + 622 in the 2 150 kev band .
the quasar is an approximate factor of 2 brighter than the average high - state flux at these energies .
previous 100 kev mev observations of the field made by osse and comptel were therefore likely dominated by the quasar emission .
the broadband 1 kev 100 mev spectrum of therefore remains uncertain .
future high - energy missions , such as _ glast _ , will have the angular resolution and flux sensitivity to both positionally associate 2cg 135 + 01 and , and detect orbital modulation in the 100 mev flux , if it exists .
we would like to thank keith jahoda and the xte gof staff for assistance with the xte software and background models .
portions of this work were funded by nasa under the rxte guest investigator program .
basic research in x - ray and radio astronomy at the naval research laboratory is supported by the office of naval research . | the binary system , , is unusual both because of the dramatic , periodic , radio outbursts , and because of its possible association with the 100 mev gamma - ray source , 2cg 135 + 01 .
we have performed simultaneous radio and _ rossi x - ray timing explorer _ x - ray observations at eleven intervals over the 26.5 day orbit , and in addition searched for variability on timescales ranging from milliseconds to hours .
we confirm the modulation of the x - ray emission on orbital timescales originally reported by taylor _
et al . _
( 1996 ) , and in addition we find a significant offset between the peak of the x - ray and radio flux .
we argue that based on these results , the most likely x - ray emission mechanism is inverse compton scattering of stellar photons off of electrons accelerated at the shock boundary between the relativistic wind of a young pulsar and the be star wind . in these observations
we also detected 2 150 kev flux from the nearby low - redshift quasar qso 0241 + 622 . comparing these measurements to previous hard x - ray and gamma - ray observations of the region containing both and qso 0241 + 622 , it is clear that emission from the qso dominates .
@xmath0 _ et al .
_ # 1to 0pt#1 |
in a previous article @xcite , referred to as paper i , we have established the general approach for the square gradient description of the interface between two phases in non - equilibrium mixtures .
we considered phenomena like temperature , density and mass fraction gradients ; heat and diffusion fluxes as well as evaporation or condensation fluxes through the interface .
some profiles were given , without going into details of the numerical procedures used to obtain them .
in this paper we will do this . in the general description of the interface one uses contributions to the helmholtz free energy density proportional to the square of the density and mass fraction gradients .
these contribution imply that it is not possible to use _ continuous local equilibrium thermodynamics _ in the interface , i.e. to calculate the local values of the various thermodynamic parameters in terms of the local density , mass fractions and temperature only .
rowlinson and widom ( see @xcite ) use the name _ point thermodynamics _ for this to distinguish it from other _ quasi_- or _ local thermodynamic _ treatments .
given the non - autonomous nature of the square gradient model , it is sensible to question whether a description in terms of excess variables along the lines given by gibbs @xcite , can be autonomous .
gibbs treatment , though only given for equilibrium systems , suggested such an assumption .
this would imply that the surface is a separate thermodynamic phase .
bakker @xcite and guggenheim @xcite made this assumption , the validity of which was subsequently disputed by defay and prigogine @xcite .
we refer to rowlinson and widom @xcite for a discussion of this point . in the theory of non - equilibrium thermodynamics of surfaces @xcite
gibbs description in terms of excess variables has been used .
it is then assumed that the gibbs description of the surface in terms of excess variables is autonomous , or in other words that one can use this property , which we will call _ local equilibrium of the surface _
, to describe the surface . in earlier work @xcite coauthored by one of us
this property was verified for one - component systems .
it is the main objective of this paper to verify this property for binary mixtures . for details about the extension of the square gradient model to non - equilibrium systems we refer to paper i. for figures of typical density , mass fraction and temperature profiles , and a discussion thereof
, we also refer to the first paper . in this paper
we focus on the properties of the excess variables .
we consider a flat interface between a binary liquid and it s vapor with the normal @xmath1 pointing from the vapor to the liquid .
we assume that all fluxes and gradients gradients are in the @xmath2direction . due to
this all variables depend only on the @xmath2coordinate .
we assume the fluid to be non - viscous , so that the viscous pressure tensor @xmath3 . in sec .
[ [ sec / equations ] ] we will give all the equations , which are required for the determination of the profiles . we shall only consider stationary states in this paper . we choose the system such that the gas is on the left hand side and the liquid is on the right hand side .
the @xmath4-axes is directed from left to right and the gravitational acceleration @xmath5 is directed towards the liquid . in sec . [
[ sec / procedure ] ] we describe numerical procedure which have been used to solve these equations .
we give the results in sec .
[ [ sec / results / profile ] ] .
we then proceed to the second part of the article , the verification of local equilibrium of the surface . in sec .
[ [ sec / localequilibrium ] ] we introduce excesses and surface variables and discuss in general the meaning of local equilibrium of the surface . in sec .
[ [ sec / results / localequilibrium ] ] we give the results of the verification procedure . finally , in sec .
[ [ sec / conclusion ] ] we give concluding remarks .
in order to calculate the various profiles in the non - equilibrium mixture a complete set of equations is required .
this set is given by the hydrodynamic balance ( conservation ) equations , the thermodynamic equations of state and the phenomenological force - flux relations .
these equations were given ( derived ) and discussed in paper i , to which we refer .
in this paper we will only list them for the stationary case . in a stationary state
the conservation equations take the following form .
the law of mass conservation is @xmath6where @xmath7 and @xmath8 are the mass density and the barycentric velocity . furthermore @xmath9 is the mass fraction of the first component and @xmath10 is the diffusion flux of the first component relative to the barycentric frame of reference .
momentum conservation is given by @xmath11where we call @xmath12 the thermodynamic tension tensor , which will be defined below
. furthermore @xmath13 and @xmath14 are the pressures parallel and perpendicular to the interface , for the planar interface under consideration . for curved surfaces
see paper i. energy conservation is given by @xmath15where @xmath16 is the total energy flux , @xmath17 is the heat flux , and @xmath18 is the total specific energy and @xmath19 is the specific internal energy .
the gradient model , discussed in the first paper @xcite gives the following expressions for the specific helmholtz energy @xmath20 , the specific internal energy @xmath19 , the parallel pressure @xmath13 , the chemical potential difference @xmath21 , the chemical potential of the second component @xmath22 and the @xmath23-element of the tension tensor @xmath24 : @xmath25 here @xmath26 is the specific helmholtz energy of the homogeneous phase , which can for instance be derived from the equation of state .
furthermore @xmath27 is the gradient contribution , where the primes indicate a derivative with respect to @xmath4 . since the equation of state
is usually given in molar quantities , it is convenient to use them here as well .
thus , @xmath28 is the molar concentration , @xmath29 is the molar fraction , where @xmath30 is the molar mass of the mixture , and @xmath31 and @xmath32 are molar masses of each component .
this energy is given by the following equation @xmath33 where the de broglie wavelength @xmath34 and the characteristic sum over internal degrees of freedom w are respectively .
@xmath35 expressions for the characteristic sums over internal degrees of freedom for each component , @xmath36 and @xmath37 , are given in paper i. in this paper they are assumed to be independent of the temperature and the molar fractions , i.e. just constant numbers .
the mixing rules for @xmath38 and @xmath39 are @xmath40 with @xmath41 , where @xmath42 as well as @xmath43 is a coefficient of a pure component @xmath44 .
we will assume in this paper that @xmath45 are independent of temperature .
this contribution is given by the following general expression for a binary mixture @xmath46 the coefficients @xmath47 , @xmath48 and @xmath49 can be expressed in the gradient coefficients @xmath50 and @xmath51 for components 1 and 2 in the following way ( see paper i for details ) @xmath52 where we use the mixing rule @xmath53 similar to the one for coefficients @xmath45 for the cross coefficient .
we will assume @xmath54 to be independent of the densities in this paper . with the above mixing rules
the gradient contribution can be written in the form @xmath55 where @xmath56 and @xmath57 , where @xmath58 .
some of the quantities from eq . can be rewritten as @xmath59 where @xmath60 , @xmath61 and @xmath62 are values of the corresponding quantities in the homogeneous phase , which are found from eq . by setting @xmath63 . for a one - component fluid @xmath64
equals the density .
for the two - component mixture @xmath64 plays a similar role as the density for the one - component fluid . we shall therefore refer to @xmath64 as the order parameter .
the size of the coefficient @xmath65 depends on the nature of the components of the mixture . in an organic mixture like cyclohexane and n - hexane , a mixture we will study in more detail in this paper , the components are very similar and as a consequence @xmath65 is small
the order parameter is then in good approximation equal to the density .
when the components are very different @xmath66 may be large and @xmath64 may become in good approximation equal to the density of one of the components ( for instance , the surface tensions of acetone and carbon disulfide are 22.7 and 31.3 gs^-2^. thus the values of corresponding @xmath67 are not very close and in their mixture the value of @xmath65 may be compared to 1 ) . in paper
i we derived the general expression for the entropy production of a mixture in the interfacial region . for a non - viscous binary mixture which has only gradients and fluxes in the @xmath2direction it takes the following form @xmath68
the resulting linear force - flux relations are : @xmath69 the resistivity coefficients @xmath70 , @xmath71 and @xmath72 will in general depend on the densities , their gradients as well as on the temperature , so they vary through the interface .
expressions for the resistivity profiles in the interfacial region are not available .
we model them , using the bulk values as the limiting value away from the surface and the order parameter profile as a modulatory curve .
@xmath73 where @xmath74 are modulatory curves for resistivity profiles . here
@xmath75 and @xmath76 are the equilibrium coexistence values of the order parameter of the gas and liquid respectively .
furthermore @xmath77 is the maximum value of the squared equilibrium order parameter gradient . for each resistivity profile
@xmath78 and @xmath79 are the equilibrium coexistence resistivities of the gas and liquid phase respectively .
coefficients @xmath80 , @xmath81 , @xmath82 control the size of the gradient term , which gives peaks in the resistivity profiles in the interfacial region .
such a peak is observed in molecular dynamic simulations of one -component fluids @xcite . limiting coefficients @xmath83 ( where @xmath84 is either @xmath5 or @xmath85 )
are related to measurable transport coefficients in the bulk phases : thermal conductivity @xmath86 , diffusion coefficient @xmath87 and soret coefficient @xmath88 . in the description of transport in the _ bulk _
phases it is convenient to use measurable heat fluxes @xmath89 where @xmath90 is a specific enthalpy of component @xmath44 in phase @xmath84 .
furthermore we used that @xmath91 . in the bulk phases
the entropy production then takes the following form : @xmath92 where we have suppressed the superscript @xmath84 for now .
the subscript @xmath93 of @xmath94 indicates that the gradient is calculated keeping the temperature constant .
using gibbs - duhem relation in a homogeneous phase at a constant pressure one can show , that @xmath95 after introducing measurable transport coefficients , the force - flux relations derived from eq . can be written in a form used in @xcite : @xmath96 comparing eq . and eq . in the bulk region
we find the bulk values @xmath97 of corresponding resistivity coefficients
@xmath98 where @xmath99 , @xmath100 .
all the quantities in eq . are taken in the specified bulk , either gas or liquid .
the numerical procedure is similar to the one , described in @xcite , however it has some differences .
we will describe the special features below .
we use the matlab procedure ` bvp4c ` @xcite to solve the stationary boundary value problem .
it requires a reasonable initial guess and boundary conditions .
we use the equilibrium profile as the initial guess .
we use a box of width 80 nm with the grid containing of 81 equidistantly spread points .
it is easier to describe equilibrium properties of the mixture using molar quantities . everywhere in this subsection
we will do this .
the superscript @xmath101 indicates a molar quantity .
the total molar concentration and molar fraction of the first component are denoted by @xmath102 and @xmath103 respectively .
equilibrium coexistence is determined by the following system of equations @xmath104 where @xmath105 , @xmath106 and @xmath107 are homogeneous chemical potentials and pressure .
@xmath108 , @xmath109 and @xmath110 , @xmath111 are coexistence density and mass fractions of gas and liquid respectively . having 6 equations eq . and 8 unknowns @xmath108 , @xmath112 , @xmath110 , @xmath111 and @xmath113 , @xmath114 , @xmath115 , @xmath116 , an equilibrium state for two - phase two component mixture contains two free parameters
particularly , the temperature and the molar fraction of the liquid phase are experimentally a reasonable choice .
we have found , however , that it is more convenient to control @xmath116 and @xmath113 in the calculations .
@xmath113 changes monotonically with @xmath112 or @xmath111 , and it is therefore a good measure for the composition . since @xmath117 , the value of @xmath118 gives the difference of the chemical potentials of two components . to obtain the equilibrium profiles @xmath119 and
@xmath120 one needs to solve a system of two differential equations @xmath121 where @xmath122 , @xmath123 and @xmath124 where @xmath31 and @xmath32 are the molar masses of the components .
this system of equations is , in fact , singular , since coefficients of the higher derivatives are proportional .
thus , we can derive one algebraic equation instead of a differential one .
@xmath125 the ` bvp4c ` procedure takes only differential equations , so we have to transform eq . to a differential one .
this can be done easily by taking derivative of both sides .
after some transformations , we have the following equation set @xmath126 where subscripts @xmath103 or @xmath102 mean partial derivative of the corresponding quantity with respect to @xmath103 or @xmath102 .
this is the system of 3 first order differential equations for 3 variables @xmath103 , @xmath127 and @xmath128 , which requires 3 boundary conditions .
one of them is eq . taken on one of the boundaries , which simply determines the integration constant for the second differential equation .
the other two are @xmath129 and @xmath130 which indicate the fact , that box boundaries are in the homogeneous region .
numerical procedure allows any values of the variables .
however , not all the values are allowed physically .
for instance , mol fraction @xmath103 is bounded in the interval @xmath131 and molar concentration is bounded in the interval @xmath132 , where @xmath39 is given in eq . .
in order to avoid out - of - range problems , we use the function which safely maps unit interval to real axes and vice versa : @xmath133 particularly , the actual variables , we provide to ` bvp4c ` procedure are @xmath134 where @xmath135 is simply the derivative of @xmath136 and @xmath137 is the limiting value for @xmath127 .
non - equilibrium conditions are implemented when we change temperature or pressure from their equilibrium values .
this results in mass and heat fluxes through the interface .
the amount of matter will then change in the gas and liquid phase .
we will put the system in such conditions , that the total contents of the box is constant and equal to the equilibrium contents .
it means , that if some amount of liquid has been evaporated , the same amount of gas is condensed externally and put back into the liquid phase .
we introduce the overall mass @xmath138 and the mass of the 1st component @xmath139 , which obey the following equations by definition @xmath140 we introduce the the overall mass flux @xmath141 , the mass flux of the 1st component @xmath142 , the energy flux @xmath143 and the `` pressure '' flux @xmath144 : @xmath145 from subsec .
[ [ sec / equations / conservation ] ] one can see , that all these fluxes are constant . from eq .
we obtain @xmath146 since @xmath147 . as in eq .
we have a singular set which leads to the algebraic equation @xmath148 taking derivative of this equation with respect to coordinate we obtain the expression for the first derivative of the fraction ( we do not give the exact expression , since it s very complicated ) @xmath149 as a consequence we have 7 unknown variables , @xmath64 , @xmath150 , @xmath151 , @xmath152 , @xmath153 , @xmath93 , @xmath94 and 4 unknown fluxes @xmath141 , @xmath142 , @xmath143 , @xmath144 .
this requires 7 first order differential equations and 11 boundary conditions ( 7 of them determine integration constants of differential equations and 4 of them determine constant fluxes ) . as differential equations we use eq . , eq . ,
one of eq . and eq . . as boundary conditions we use the following :
the first boundary condition is eq .
taken on one of the boundaries , which simply determine the integration constant for eq . .
the 4 other conditions control the overall content of the box ( particularly , it is the same as in equilibrium ) @xmath154 here @xmath155 and @xmath156 are the equilibrium values of the total overall mass and the total mass of the 1st component in the whole box .
the 2 more conditions are @xmath157 which indicate the fact , that box boundaries are in the homogeneous region . in contrast to the equilibrium case
, the density in the non - equilibrium homogeneous region may vary with coordinate , so @xmath150 differs from zero on the boundaries , and , in fact , they do .
the value of the @xmath150 is however small in the homogeneous region , comparing to the value in the surface region , so we may neglect it and use such approximation .
this will lead to the wrong profile behavior only in the small vicinity on the boundary , which we will exclude from the further analysis .
the 4 last boundary conditions reflect the conditions of the probable experiment .
for instance , we may control the temperatures on the both side of the box , the pressure of the vapor side and the fraction on the liquid side . @xmath158 to solve these equations numerically we use the same technics as in the equilibrium case .
all the variables should be properly scaled in order to make them to be the same order of magnitude .
this balances the numerical residual and gives better solution result .
we use the following variables @xmath159 where @xmath160 and scaling parameters @xmath161 , @xmath162 , @xmath163 , @xmath164 .
in this section we show some profiles , obtained with the help of the above procedure .
we choose a mixture of hexane and cyclohexane and give some of their properties relevant for our calculation in the table below .
we note , that among them only the molar masses have been measured .
there are number of problems to obtain the values of other material properties .
we determined , for instance , the two van der waals coefficients of the pure phases using their critical temperatures and pressures and @xmath165 were used . ] .
as a consequence the critical volumes per mole found in our description for the pure components differs substantially from the experimental value . for the mixture
the van der waals coefficients were then found using the mixing rules .
we use the values of the molar mass and the van der waals coefficients given in table [ [ tbl / const / material ] ] : .the molar mass and the van der waals coefficients [ cols="<,<,<,<",options="header " , ] together with the `` mixing '' rules for the heat conductivity @xmath166 the values of the gradient coefficients are not available at all .
one can determine them comparing the actual value of the surface tension of a pure fluid with the one , calculated with a given @xmath54 .
for given conditions the value of the surface tension of the mixture is about 0.027 n / m . we therefore choose @xmath167 to be equal @xmath168 j m^5^/mol^2^ and @xmath169 .
this gives values of the surface tension about 0.03 n / m .
the equilibrium properties of the system are calculated in the following sections .
] at @xmath170 k and @xmath171 j / mol .
this gives @xmath172 pa , @xmath173 j / mol , @xmath174 mol / m^3^ , @xmath175 mol / m^3^ , @xmath176 and @xmath177 .
the mixture is then perturbed from equilibrium for the following three cases : 1 ) setting @xmath178 equal to 0.98 , 0.99 , 1.01 , 1.02 of @xmath116 and keeping @xmath179 , @xmath180 and @xmath181 equal to their equilibrium values , see fig . [
[ fig_tl ] ] ; 2 ) setting @xmath180 equal to 0.98 , 0.99 , 1.01 , 1.02 of @xmath115 and keeping @xmath179 , @xmath178 and @xmath181 equal to their equilibrium values , see fig .
[ [ fig_pg ] ] ; 3 ) setting @xmath181 equal to 0.98 , 0.99 , 1.01 , 1.02 of @xmath182 and keeping @xmath179 , @xmath178 and @xmath180 equal to their equilibrium values , see fig .
[ [ fig_zil ] ] . ,
title="fig:"][fig_t_tl ] , title="fig : " ] [ fig_psi_tl ] , title="fig:"][fig_t_pg ] , title="fig : " ] [ fig_psi_pg ] , title="fig:"][fig_t_zil ] , title="fig : " ] [ fig_psi_zil ]
in equilibrium it is possible to describe the surface in terms of gibbs excess quantities @xcite .
one can treat a system of coexisting liquid and vapor as a three - phase system : liquid and vapor bulk phases and the surface phase .
the surface has thermodynamic properties .
the temperature and chemical potentials have the same equilibrium value as in the rest of the system .
furthermore the thermodynamic state of the surface is given by excess concentrations and thermodynamic potentials .
following gibbs we have for the surface @xmath183 the superscript @xmath184 indicates here the surface and equal to the excess of corresponding quantities . in these relations the temperature and the chemical potentials , which are the same everywhere , are independent of the choice of the dividing surface .
the excesses depend on the choice of the dividing surface , in a such way that the above relations are true for any choice of the dividing surface .
it is our aim in this paper to show that the surface in a non - equilibrium liquid - vapor system can also be described as a separate thermodynamic phase using the gibbs excess quantities .
we will call this property the local equilibrium of the surface .
the property of local equilibrium for the surface implies that it is possible to define all thermodynamic properties of a surface such that they have their equilibrium coexistence values for any choice of the dividing surface given the temperature of the surface @xmath185 and the chemical potential difference @xmath186 . for this purpose
we will define the excess densities and develop a method to obtain @xmath185 and @xmath186 independent of the choice of the dividing surface below .
the analysis will be done using the numerical solution of the system in stationary non - equilibrium states .
the definition of the excesses consists of 3 steps : determining the phase boundaries , defining the specified dividing surface and , in particular , defining the excesses . to determine the phase boundaries we will use the order parameter @xmath64 .
we introduce a small parameter @xmath187 and define the @xmath187-dependent boundary , @xmath188 between the vapor and the surface by @xmath189 where @xmath190 is the equation of state s value ( no gradient contributions ) of @xmath64 for pressure @xmath191 , mass fraction @xmath153 and temperature @xmath93 . the @xmath187-dependent boundary , @xmath192 , between the surface and the liquid is defined in the same way
. the numerical procedure calculates profiles only at specified grid points , which we provide to the procedure .
that means that @xmath188 and @xmath192 can only be situated at points of the grid .
we choose their position to be the last bulk point of the grid where the left hand side of eq .
does not exceed the right hand side . in our calculations
we will choose @xmath193 and use a grid of 81 points .
we shall also choose bulk boundaries near the box boundary where , because of the finite size of the box , the behavior of the profiles might be uncharacteristic . to avoid this effect , we do not consider the first 5 points of each phase close to these boundaries when we calculate the properties in these phases . the 6th point we call @xmath194 and the 76th point @xmath195 .
the bulk gas therefore ranges from @xmath194 to @xmath188 and the bulk liquid ranges from @xmath192 to @xmath195 .
the surface therefore ranges from @xmath188 to @xmath192 . in order to define excess quantities
properly we always choose conditions such , that the bulk widths , @xmath196 and @xmath197 are larger then the surface width @xmath198 . in order to determine excess densities we need to extrapolate bulk profiles into the interfacial region .
in equilibrium extrapolated bulk profiles are constants which are equal to the coexisting values of the corresponding quantities .
non - equilibrium bulk profiles are not constant .
we fit the bulk profile with a polynomial of order @xmath199 and use this polynomial to extrapolate non - equilibrium bulk profiles into the interfacial region .
this is done with the help of matlab functions ` polyfit ` and ` polyval ` .
it is important to realize that the extrapolation of the bulk profiles introduces a certain error depending on the choice of @xmath187 and @xmath200 in particular for non - equilibrium systems .
the distances between commonly used dividing surfaces , such as for instance the equimolar surface and the surface of tension , are very small . thus , if there occurs an error in the determination of a dividing surface using a course grid , that would lead to inaccurate results .
we therefore divide each interval of the course grid between @xmath188 and @xmath192 in @xmath201 subintervals .
this surface grid is used for all operations related to the surface . within the interfacial region
we interpolate all the profiles ( which were obtained by extrapolation from the bulk to the surface region using the course grid ) from the course grid to the surface grid using a polynomial of order @xmath202 with the help of matlab functions ` polyfit ` and ` polyval ` .
we can now define the excess @xmath203 of any density @xmath204 as a function of a dividing surface @xmath205 as @xmath206}\ ] ] where @xmath207 and @xmath208 are the extrapolated gas and liquid profiles and @xmath209 is the heaviside function .
the density @xmath210 is per unit of volume and @xmath203 is per unit of surface area . in our calculations integration
is performed using the trapezoidal method by matlab function ` trapz ` .
we can now define different dividing surfaces .
the equimolar dividing surface @xmath211 is defined by the equation @xmath212 .
analogously we define equimolar surfaces with respect to component 1 and 2 : @xmath213 and @xmath214 , and the equidensity surface @xmath215 .
the surface of tension @xmath216 is defined from the equation @xmath217 .
all the densities are given as arrays on a coordinate grid , but not as continuous functions .
thus , in order to find the solution of an equation @xmath218 we calculate the values @xmath219 for each point @xmath220 within the surface region and find the minimum of it s absolute value , @xmath221 . because of the discrete nature of the argument this value may not be equal to zero , but it will be the closest to zero among all other coordinate points .
so we will call this point the root of the equation @xmath222 .
we use the fine surface grid in this procedure . if follows from eq .
that @xmath223 which we will use later .
an equilibrium two - phase two - component mixture has two free parameters , for instance the temperature @xmath93 and the chemical potential difference @xmath224 .
_ local equilibrium of a surface
_ implies , that in non - equilibrium it should be possible to define the temperature @xmath185 and the chemical potential difference @xmath186 of the surface . as found in ref .
@xcite for the surface temperature in the one - component system , both @xmath185 and @xmath225 should be independent of the choice of the dividing surface .
the equilibrium temperature and chemical potential difference determine all other equilibrium properties of the surface .
thus , there is a bijection from @xmath116 and @xmath226 to any other set of independent excess variables @xmath227 and @xmath228 , so that one can use them equally well in order to characterize a surface . in non - equilibrium
the temperature and chemical potential difference vary through the interfacial region , but as @xmath229 and @xmath230 are excesses , they characterize the whole surface .
if a non - equilibrium surface is in local equilibrium , there should exist the same bijection .
this implies that given two independent non - equilibrium excesses @xmath229 and @xmath230 one can determine the temperature @xmath185 and the chemical potential @xmath186 of the whole surface .
thus one can calculate equilibrium tables of @xmath227(@xmath116 , @xmath226 ) and @xmath228(@xmath116 , @xmath226 ) for different values of @xmath116 and @xmath226 and then determine temperature and chemical potential of a surface as @xmath231(@xmath229 , @xmath230 ) and @xmath232(@xmath229 , @xmath230 ) .
as we want the temperature and chemical potential difference to be independent of the position of the dividing surface , we shall use excesses which are also independent of the position of a dividing surface in equilibrium for @xmath233 and @xmath234 . for two component mixture
these independent variables are the surface tension @xmath235 and the relative adsorption @xmath236 .
if the number of components is more then 2 , additional relative adsorptions should be used .
these quantities are well defined for equilibrium , but not for non - equilibrium .
so we will define them first . in equilibrium
the surface tension is defined as minus the excess of the parallel pressure @xmath237 .
alternatively one often uses the integral of @xmath238 across the interface : @xmath239 .
both definitions are equivalent in equilibrium since @xmath191 is constant through the interface and @xmath240 is identically zero in the bulk phases . in non - equilibrium @xmath240
may differ from zero in the bulk regions , however , and this makes the second definition inappropriate .
we will therefore define the non - equilibrium surface tension using the standard definition @xmath241 this quantity differs from @xmath242 by the term equal to @xmath243 , which is usually small compared to @xmath244 .
the relative adsorption is defined as @xmath245 in equilibrium @xcite , where @xmath246 and @xmath247 are coexistence concentrations of the corresponding components .
since these quantities are not constant in non - equilibrium , we can not use this definition directly .
one can however see from eq .
, that both , in equilibrium and non - equilibrium @xmath248 , where the prime indicates a spatial derivative .
since in equilibrium @xmath249 we can use the following definition @xmath250 both in equilibrium and non - equilibrium .
if the system is in local equilibrium we may write : @xmath251 substituting the expressions for @xmath252 and @xmath253 from eq .
and eq . into eq . we obtain the following relations @xmath254 this gives the bijection equations to determine @xmath185 and @xmath186 from the actual non - equilibrium variables . as the left hand sides in eq
are in good approximation independent of the position of the dividing surface , @xmath185 and @xmath186 are similarly independent on this position .
the other quantities required for the gibbs description of the non - equilibrium surfaces we define in the following way .
the surface chemical potentials are the equilibrium coexistence values determined via the procedure discussed in subsec .
[ [ sec / procedure / equilibrium ] ] @xmath255 we define the surface extensive properties as @xmath256 the local equilibrium of a surface should be established for any choice of a dividing surface .
the results of the calculations for any particular choice of a dividing surface may not be representative since they may be different differ for another choice of a dividing surface .
thus , the property of local equilibrium should be established for all dividing surfaces together .
k and @xmath257 j / mol .
the vertical lines indicate the @xmath258 , @xmath216 , @xmath211 , @xmath259 dividing surfaces from left to right . ] .
the vertical lines indicate the @xmath258 , @xmath216 , @xmath211 , @xmath259 dividing surfaces from left to right . ] .
the vertical lines indicate the surface boundaries and @xmath258 , @xmath216 , @xmath211 , @xmath259 dividing surfaces . ]
consider the profile of an excess @xmath260 as a function of position of a dividing surface @xmath205 .
it follows from eq . that the slope of the excess profile is equal to the difference between the extrapolated values of a profile of thermodynamic quantity @xmath210 . in equilibrium
these values are constant and equal to the coexistence values .
thus , equilibrium excess profiles are straight lines , as one can see on fig .
[ [ eq_excesses ] ] .
non - equilibrium profiles in the bulk phases are not constant .
we construct the extrapolated profiles using @xmath200th order polynomials with @xmath199 .
resulting non - equilibrium excesses are therefore polynomials of the order @xmath261 , according to eq . .
these profiles , for the most extreme case of non - equilibrium perturbation @xmath262 , are shown in fig .
[ [ noneqtl102_excesses ] ] . even though these profiles are polynomials of the 3rd order they are very close to straight lines .
as one can see from fig .
[ [ noneqtl102_diff_excesses ] ] the variation in the slope is about 1% through the whole surface .
it indicates that this non - equilibrium `` state '' is very close to an equilibrium one .
we therefore develop the procedure to relate the non - equilibrium state to an equilibrium one by comparing thermodynamic quantities in equilibrium and in non - equilibrium .
the comparison performed in one particular point of the surface may not be sufficient because it may suffer from artefacts peculiar to this particular surface .
moreover , any comparison performed in a particular point does not speak for the whole .
we therefore compare the non - equilibrium surface with an equilibrium one for all dividing surfaces together .
we will use the least square sum method for this . consider
a non - equilibrium thermodynamic excess @xmath263 and a quantity @xmath264 which is a combination of excesses and may depend on @xmath265 as parameters .
we introduce the following measures of the difference of @xmath266 and @xmath267 @xmath268 and @xmath269^{2 } } \\ \\
\sigma_{t^{{\scriptstyle}s},r^{{\scriptstyle}s}}(t , \psi ) & \equiv \displaystyle { 1 \over n}\,\sqrt{s_{t^{{\scriptstyle}s},r^{{\scriptstyle}s}}(t , \psi ) } \end{array}\ ] ] where @xmath270 is the number of surface points .
we say that @xmath266 and @xmath267 are _ the same _ in the surface if the value of @xmath271 is negligible compared to the typical value of either @xmath272 or @xmath273 .
if two quantities @xmath266 and @xmath267 are the same in the above sense , we say that the non - equilibrium state of the surface is characterized by surface temperature @xmath274 and chemical potential difference @xmath275 if @xmath276 here superscript @xmath184 indicates that we speak about surface quantities only ( as everywhere in this paper ) and subscript @xmath277 indicates that @xmath274 and @xmath275 are the parameters for all dividing surfaces together ( in contrast to the values @xmath278 and @xmath279 determined from eq . for each particular dividing surface @xmath205 ) .
these definitions are easy to illustrate in equilibrium .
for instance , for @xmath280 it follows from eq .
, that @xmath281 . furthermore @xmath282 .
thus @xmath283 and @xmath284 .
it is also true that @xmath285 and @xmath286 .
according to the above definitions i ) @xmath287 and @xmath288 are the same quantities , but @xmath287 and @xmath289 are not the same ; ii ) the equilibrium state is characterized by @xmath290 ; as it should be . while this analysis is trivial in equilibrium , it is not trivial in non - equilibrium . note , that while in equilibrium the conditions @xmath291 and @xmath292 are equivalent , in general it does not follow in non - equilibrium from eq .
that @xmath293 thus eq .
is not a good measure of the equality of the quantities and states in non - equilibrium .
we may therefore speak about the equality of thermodynamic quantities as well as about the state @xmath185 and @xmath186 of the surface in non - equilibrium only in the least square sense , as it is given in eq . .
within establishing the local equilibrium property of a non - equilibrium surface we would like to verify the following properties : i ) the existence of the unique temperature @xmath185 and chemical potential difference @xmath186 of a non - equilibrium surface ; ii ) the validity of the eq . in non - equilibrium at the surface
s @xmath185 and @xmath186 ; iii ) the possibility to determine all the properties of a non - equilibrium surface from equilibrium tables at the surface s @xmath185 and @xmath186 .
we do this in the following section .
we calculate the the equilibrium properties ( coexistence data , such as the pressure or bulk densities , as well as various excesses ) of the system for the range of temperatures @xmath294 k and the range of chemical potentials @xmath295 j / mol .
the value of a thermodynamic quantity at any point @xmath265 , which is between these is interpolated using the matlab procedures ` interp2 ` and ` griddata ` .
as was mentioned , in equilibrium both @xmath296 and @xmath236 are independent of the location of the dividing surface @xmath205 . given the above definitions , eq .
and eq . , we can calculate these quantities for non - equilibrium states .
calculations show , that even though @xmath235 and @xmath236 are not exactly independent on @xmath205 away from equilibrium , the relative deviation is so small ( about @xmath297 for @xmath235 and @xmath298 for @xmath236 in the worst case ) , that one can consider these quantities to be independent of the position of the dividing surface .
thus one may use them in order to find the temperature , @xmath185 , and the chemical potential difference , @xmath186 , of the surface in non - equilibrium states , which will be independent of the position of the dividing surface . using eq . and eq . together with eq .
we construct the following expressions @xmath299^{2 } } \\ \\
s_{\gamma_{12}}(t,\psi ) & = \displaystyle\sum_{x^{\scriptstyle s}}{\left [ { \frac{\widehat{c}^{\;\prime}_{1,eq}({x^{\scriptstyle s } } ; t,\psi)}{\widehat{c}^{\;\prime}_{1}({x^{\scriptstyle s } } ) } } -{\frac{\widehat{c}^{\;\prime}_{2,eq}({x^{\scriptstyle s } } ; t,\psi)}{\widehat{c}^{\;\prime}_{2}({x^{\scriptstyle s } } ) } } \right]^{2 } } \\ \end{array}\ ] ] where the prime indicates the derivative with respect to @xmath205 . for @xmath300 being the profile of the surface tension @xmath235 for the case of perturbing @xmath262 .
the lines in the @xmath93-@xmath94 plane are lines of constant value of @xmath301 @xmath302 ( where @xmath44 is either @xmath235 or @xmath236 ) should reach the minimum at @xmath274 and @xmath275 .
we note however , that neither @xmath302 have a minimum at a single point @xmath303 .
there is a whole generatrix curve of minima @xmath304 so the plot of @xmath302 is a valley .
one can see it on fig .
[ [ noneqtl102_emc ] ] .
every point of the generatrix curve is the minimum point of @xmath302 along the direction `` perpendicular '' to this generatrix . if @xmath305 or @xmath306 and @xmath307 represent the corresponding profiles for some equilibrium state @xmath290 , then @xmath308 and the generatrix is constant . since these profiles are non - equilibrium profiles , the generatrix is not exactly constant but very close to it
. thus @xmath309 , where @xmath310 is a direction in @xmath93-@xmath94 plane which is perpendicular to generatrix .
in fact , one should be careful speaking about directions , since no metric is defined in the @xmath93-@xmath94 plane .
thus we can not introduce @xmath311 so that @xmath312 .
in fact , @xmath310 can be any direction which does not coincide or does not almost coincide with the direction of generatrix . in practice
we find that we can use @xmath313 , while using @xmath314 gives less accurate results .
thus we determine the minima curve from the equation @xmath315 thus one needs two quantities @xmath316 and @xmath317 in order to determine @xmath274 and @xmath275 uniquely .
the surface temperature and chemical potential difference @xmath274 and @xmath275 are determined from the intersection of two minimum curves of @xmath316 and @xmath317 @xmath318 we calculate the temperatures and the chemical potential differences for different non - equilibrium conditions .
they are outlined in tables [ [ tbl / tp / tl]-[tbl / tp / zil ] ] .
the first row of each table , corresponding to @xmath277 , gives @xmath185 and @xmath186 calculated from eq . .
the following rows give , corresponding to different particular dividing surfaces , gives @xmath185 and @xmath186 calculated from eq . .
l@l@l@l@l@ + & & + surface & @xmath185 & @xmath186 & @xmath185 & @xmath186 + @xmath277 & 331.831 & 770.53 & 328.129 & 650.92 + & & & & + @xmath211 & 331.823 & 769.51 & 328.124 & 650.29 + @xmath216 & 331.828 & 770.22 & 328.123 & 650.21 + @xmath259 & 331.814 & 767.97 & 328.127 & 650.43 + @xmath258 & 331.838 & 771.86 & 328.121 & 650.1 + l@l@l@l@l@ + & & + surface & @xmath185 & @xmath186 & @xmath185 & @xmath186 + @xmath277 & 330.796 & 683.87 & 329.059 & 696.52 + & & & & + @xmath211 & 330.8 & 684.68 & 329.063 & 697.22 + @xmath216 & 330.799 & 684.44 & 329.063 & 697.12 + @xmath259 & 330.804 & 685.19 & 329.065 & 697.46 + @xmath258 & 330.795 & 683.93 & 329.061 & 696.87 + l@l@l@l@l@ + & & + surface & @xmath185 & @xmath186 & @xmath185 & @xmath186 + @xmath277 & 329.577 & 559.32 & 330.24 & 812.86 + & & & & + @xmath211 & 329.598 & 562.44 & 330.242 & 813.16 + @xmath216 & 329.584 & 560.5 & 330.253 & 814.67 + @xmath259 & 329.63 & 566.83 & 330.219 & 809.99 + @xmath258 & 329.554 & 556.3 & 330.278 & 817.99 + note , that @xmath185 and @xmath186 may be different from the continuous values in the interfacial region . in this section
we would like to verify that the surface quantities defined by eq .
satisfy eq . with @xmath185 and @xmath186 determined by eq .
and eq . .
@xmath319 namely , with the definition eq . ,
@xmath320 where the right hand side is @xmath321 of the corresponding quantity .
eq . is the non - equilibrium analogon of equilibrium eq . .
in order to analyze the measure of validity of eq .
we construct the quantity @xmath322^{2 } } \\ \\
\displaystyle \epsilon_{\phi_{gibbs}}({x^{\scriptstyle s } } ; t , \psi ) & = \displaystyle \left| \frac{\phi^{s}({x^{\scriptstyle s}})-\phi^{s}_{gibbs}({x^{\scriptstyle s } } ; t , \psi)}{\phi^{s}({x^{\scriptstyle s}})}\right| \end{array}\ ] ] for each thermodynamic potential @xmath323 , @xmath19 , @xmath20 , @xmath5 .
@xmath324 gives the relative error of the determination of the surface quantity @xmath325 using the gibbs excesses relations eq .
for for all dividing surfaces together , while @xmath326 gives this error for particular dividing surface .
we build @xmath327 for @xmath328 , @xmath329 determined from eq . only for the whole surface .
we build @xmath330 both for @xmath328 , @xmath329 determined for the whole surface and for @xmath331 , @xmath332 determined from eq . for particular dividing surface .
the values of the corresponding errors are listed in tables [ [ tbl / gibbs / tl]-[tbl / gibbs / zil ] ] in subsec .
[ [ sec / excesserrors / gibbs ] ] , and are found to be small . as one can see , there is a variation in the value of the error for the different dividing surfaces .
it is caused by two reasons .
the first reason for this is slight variation in @xmath185 and @xmath186 from tables [ [ tbl / tp / tl]-[tbl / tp / zil ] ] for different dividing surfaces .
the variation of each excess potential corresponds to the variation of @xmath185 and @xmath186 through these surfaces .
thus so do the relative errors .
another factor which influences the value of these errors is the actual value of an excess at a given dividing surface .
if it is close to zero , then in the expression for @xmath333 the small value is in denominator and it gives the huge value for the error .
particularly , @xmath334 both in equilibrium and in non - equilibrium which makes the row corresponding to @xmath5 at @xmath211 be uninformative and one should not take into account these data . for the case of perturbing @xmath262 .
the vertical lines indicate the surface boundaries and @xmath335 dividing surfaces . ]
we emphasize however that the overall error @xmath324 represents the whole surface and thus do not suffer from the fact that some quantity is negligible at some dividing surface .
there are such points for each potential @xmath210 , however their contribution to the whole error is negligible itself .
so we can see , that if the particular dividing surface is far from zero point of @xmath210 , @xmath326 gives the good measure of the error .
while if the particular dividing surface is close to zero point of @xmath210 , @xmath326 fails to measure the error .
one can see from fig .
[ [ noneqtl102_epsilon_gibbs ] ] that the relative error @xmath326 indeed rises enormously at @xmath336 .
particularly because of this fact the definition of the excess quantities in @xcite was different from eq . .
another possible test is to compare the absolute error @xmath337 with the deviation @xmath338 defined in eq . .
the calculations show that for the particular dividing surfaces the former quantity does not exceed the lateral , which indicates that all the absolute errors are actually within the trust region .
in this subsection we will verify the possibility to determine all the properties of a non - equilibrium surface from equilibrium tables at the surface s @xmath185 and @xmath186 .
the surface chemical potentials @xmath339 and @xmath340 are already defined as their equilibrium values by eq . .
so in this section we will verify the relation @xmath341 as in subsec .
[ [ sec / results / localequilibrium / gibbs ] ] we compare the actual excess of a thermodynamic potential with the corresponding equilibrium value at given temperature and chemical potential of the surface .
before we do this a note has to be made . under non - equilibrium conditions the profile of a quantity @xmath325
is shifted with respect to the equilibrium one .
one can see this in the example for @xmath342 in fig .
[ [ noneqtl102_table_excesses ] ] . the reason for this
are fluxes caused by non - equilibrium perturbation .
the whole surface is therefore shifted .
one can clearly see that comparing the positions of the particular dividing surfaces on fig .
[ [ eq_excesses ] ] and fig .
[ [ noneqtl102_excesses ] ] .
so the direct comparison of the profiles should be done not in the observer s frame of reference ( ofo , which is used in all other calculations ) , but in the _ surface s frame of reference _ ( sfo ) .
the sfo is simply shifted with respect to the ofo , depending on the rate of non - equilibrium perturbations .
zero of the sfo is chosen at the reference surface , which can be either the equimolar surface , or any other physically sensible surface .
if @xmath343 is the position of this surface in ofo and @xmath344 is the profile of @xmath325 in ofo , then @xmath345 is the profile of @xmath325 in sfo . for the case of perturbing @xmath262 ( solid line )
compared to the profiles of @xmath288 calculated from the equilibrium tables for different @xmath116 and @xmath226 ( dotted lines ) .
the vertical lines indicate the surface boundaries and dividing surfaces for non - equilibrium case.,title="fig:"][noneqtl102_table_excesses1 ] for the case of perturbing @xmath262 ( solid line ) compared to the profiles of @xmath288 calculated from the equilibrium tables for different @xmath116 and @xmath226 ( dotted lines ) .
the vertical lines indicate the surface boundaries and dividing surfaces for non - equilibrium case.,title="fig : " ] [ noneqtl102_table_excesses2 ] we can now determine to which equilibrium state the non - equilibrium one should correspond . consider the following definitions of @xmath346 and @xmath347 which have the same meaning as in eq .
: @xmath348^{2 } } \\ \\
\displaystyle \epsilon_{\phi_{table}}({x^{\scriptstyle s } } ; t , \psi ) & = \displaystyle \left| \frac{\phi^{s}({x^{\scriptstyle s}})-\phi^{s}_{eq}({x^{\scriptstyle s } } ; t , \psi)}{\phi^{s}({x^{\scriptstyle s}})}\right| \end{array}\ ] ] for each thermodynamic potential @xmath323 , @xmath19 , @xmath20 , @xmath5 . @xmath343 and @xmath349 are the non - equilibrium and equilibrium positions of the reference surface in ofo .
the set @xmath350 is the non - equilibrium surface grid and is used for both profiles .
since the width of an equilibrium surface may be not the same as the non - equilibrium one , the summation may exceed the formal boundaries of the equilibrium surface .
this is not a problem however , since the equilibrium profile @xmath351 is the line with constant slope everywhere , as well as beyond the formal boundaries .
we do nt shift the surface grid in the definition of @xmath352 because in that notation @xmath205 means the particular dividing surface , while @xmath353 means the point of the surface grid .
the values of the corresponding errors are listed in tables [ [ tbl / table / tl]-[tbl / table / zil ] ] in subsec .
[ [ sec / excesserrors / table ] ] , and are found to be small . as in subsec .
[ [ sec / results / localequilibrium / gibbs ] ] we see , that for the equimolar surface the relative error in @xmath5 is huge . there is the same reason for this , namely that @xmath354 both in equilibrium and in non - equilibrium .
this again makes the row corresponding to @xmath5 at @xmath211 be uninformative and one should not take into account these data .
the article continues the general analysis started in @xcite . here
we focus on a specified mixture and develop further its properties .
we choose a binary mixture of cyclohexane and @xmath0-hexane as the system and describe in details how to implement the general analysis presented in @xcite .
we build numerical procedure for solving the resulted system of differential equations .
the resulted profiles of continuous variables are presented in sec .
[ [ sec / results / profile ] ] and in the first article .
we see , in particular , that a two component mixture develop the temperature profile in the surface region which is similar to the temperature profile obtained for one - component system @xcite .
another characteristic of a binary mixture is the difference between chemical potentials of components .
the behavior of the profile of @xmath94 in non - equilibrium steady - states is different .
it has different values in the different bulk phases and we observe a transition from the one value to the other in the surface region . we then proceed to verify the local equilibrium of the surface .
this property means that a surface under non - equilibrium steady - state conditions can be described as equilibrium one in terms of the gibbs excess densities .
we have discussed the meaning of the surface quantities in non - equilibrium and established the systematic procedure to obtain them .
we were in particular focused on i ) the existence of the unique surface s temperature and chemical potential difference ; ii ) the validity of the relations between thermodynamic gibbs excesses in a non - equilibrium surface ; iii ) the correspondence between the non - equilibrium and equilibrium properties of the surface .
it was possible to verify that one can speak about these statements independently on the choice of the dividing surface .
similar results for the one - component system were obtained in @xcite .
the extrapolations procedure is numerical , and contains therefore a certain error .
we may not expect this error to be negligible , not only because of numerical inaccuracy , but also because of the non - equilibrium nature of the system .
if the error is within a reasonable range , we will consider this as a satisfactory verification of local equilibrium .
the main part of the analysis in the interfacial region is introducing the excesses of thermodynamic densities , which are constructed with the help of extrapolated bulk profiles .
in contrast to equilibrium , non - equilibrium bulk profiles are not constants , and therefore their extrapolation to the surface region is not always accurate .
the accuracy of extrapolation lowers when the surface width increases .
apparent small deviations from local equilibrium are therefore to some extent an artefact of the inaccuracy of the extrapolation . in the description of the surface
excess densities it may happen that for particular choice of the dividing surface not one but several of the excesses are negligible .
this increases the relative error enormously while the absolute error remains finite and more or less constant . in order to avoid this problem
we consider the excesses for all dividing surfaces together , rather then for a particular dividing surface .
particularly , in @xcite the definition of the excess gibbs energy was chosen differently because this excess was very small for the equimolar surface .
we have shown in this paper why this is not needed .
one can see from these data , that within different ways of perturbing mixture from equilibrium the biggest error comes when one perturbs the temperature on the liquid side .
this is the most extreme condition for the mixture being in non - equilibrium .
while the relative temperature perturbation is only 2% , the resulting temperature gradient is about @xmath355 k / m which is very far beyond ordinary non - equilibrium conditions .
the other perturbations make the validity of local equilibrium for the surface more precise .
similarly smaller perturbations make the validity of local equilibrium also more precise .
we therefore conclude that the local equilibrium of the surface is valid with a reasonable accuracy .
+ & & & + @xmath210 & error & for @xmath356 & for @xmath357 & for @xmath356 & for @xmath357 + @xmath323 & @xmath358 & 0.01328 & - & 0.033276 & - + & & & & & + & @xmath359 & 0.023799 & 0.023419 & 0.064597 & 0.065002 + & @xmath360 & 0.020605 & 0.020737 & 0.056585 & 0.057148 + & @xmath361 & 0.035878 & 0.033571 & 0.085212 & 0.085204 + & @xmath362 & 0.015606 & 0.016541 & 0.047847 & 0.048578 + + @xmath19 & @xmath358 & 0.0071257 & - & 0.026714 & - + & & & & & + & @xmath359 & 0.016729 & 0.016462 & 0.045109 & 0.045392 + & @xmath360 & 0.015059 & 0.015156 & 0.041048 & 0.041457 + & @xmath361 & 0.022033 & 0.020616 & 0.054286 & 0.05428 + & @xmath362 & 0.012161 & 0.012889 & 0.036244 & 0.036797 + + @xmath20 & @xmath358 & 0.20039 & - & 0.12983 & - + & & & & & + & @xmath359 & 7.3727 & 7.3666 & 8.7959 & 8.799 + & @xmath360 & 1.9809 & 1.981 & 2.2605 & 2.2602 + & @xmath361 & 1.6966 & 1.6881 & 2.5586 & 2.5602 + & @xmath362 & 0.65429 & 0.65354 & 1.0109 & 1.0095 + + @xmath5 & @xmath358 & 0.36323 & - & 0.15487 & - + & & & & & + & @xmath359 & 272.37 & 272.15 & 73.468 & 73.494 + & @xmath360 & 2.7241 & 2.7242 & 3.2313 & 3.2309 + & @xmath361 & 1.4054 & 1.3983 & 1.9592 & 1.9605 + & @xmath362 & 0.72857 & 0.72773 & 1.1818 & 1.1802 + + + & & & + @xmath210 & error & for @xmath356 & for @xmath357 & for @xmath356 & for @xmath357 + @xmath323 & @xmath358 & 0.0013703 & - & 0.0019532 & - + & & & & & + & @xmath359 & 0.0018964 & 0.0022352 & 0.0011722 & 0.00088793 + & @xmath360 & 0.0017063 & 0.0019978 & 0.0010683 & 0.00083823 + & @xmath361 & 0.0025175 & 0.0030049 & 0.0015513 & 0.0010756 + & @xmath362 & 0.0014259 & 0.0016451 & 0.00090372 & 0.00076269 + + @xmath19 & @xmath358 & 0.00041867 & - & 0.00033213 & - + & & & & & + & @xmath359 & 0.0013316 & 0.0015695 & 0.00081921 & 0.00062057 + & @xmath360 & 0.00124 & 0.001452 & 0.00077388 & 0.00060722 + & @xmath361 & 0.0015926 & 0.001901 & 0.00096161 & 0.00066669 + & @xmath362 & 0.001093 & 0.001261 & 0.00069529 & 0.00058679 + + @xmath20 & @xmath358 & 0.043422 & - & 0.033692 & - + & & & & & + & @xmath359 & 1.4194 & 1.4152 & 1.2845 & 1.2884 + & @xmath360 & 0.3884 & 0.38805 & 0.37919 & 0.3795 + & @xmath361 & 0.40171 & 0.39739 & 0.30303 & 0.30621 + & @xmath362 & 0.13128 & 0.1311 & 0.13976 & 0.13952 + + @xmath5 & @xmath358 & 0.08442 & - & 0.021892 & - + & & & & & + & @xmath359 & 28.236 & 28.152 & 48.786 & 48.935 + & @xmath360 & 0.56258 & 0.56208 & 0.54645 & 0.5469 + & @xmath361 & 0.3187 & 0.31528 & 0.24447 & 0.24704 + & @xmath362 & 0.15021 & 0.15001 & 0.1583 & 0.15802 + + + & & & + @xmath210 & error & for @xmath356 & for @xmath357 & for @xmath356 & for @xmath357 + @xmath323 & @xmath358 & 0.00035435 & - & 0.0011482 & - + & & & & & + & @xmath359 & 0.0023134 & 0.0028072 & 0.0042669 & 0.0042419 + & @xmath360 & 0.0021873 & 0.0026065 & 0.0039043 & 0.0039484 + & @xmath361 & 0.0028054 & 0.0035457 & 0.0054811 & 0.0052479 + & @xmath362 & 0.0020102 & 0.0023063 & 0.0033597 & 0.0035215 + + @xmath19 & @xmath358 & 0.00040969 & - & 0.0015073 & - + & & & & & + & @xmath359 & 0.0016206 & 0.0019665 & 0.0029893 & 0.0029717 + & @xmath360 & 0.0015861 & 0.0018901 & 0.0028345 & 0.0028665 + & @xmath361 & 0.0017517 & 0.0022139 & 0.0034446 & 0.003298 + & @xmath362 & 0.0015408 & 0.0017677 & 0.0025842 & 0.0027086 + + @xmath20 & @xmath358 & 0.0063873 & - & 0.0026805 & - + & & & & & + & @xmath359 & 0.17101 & 0.15419 & 0.033343 & 0.034999 + & @xmath360 & 0.014471 & 0.015093 & 0.023261 & 0.022844 + & @xmath361 & 0.15718 & 0.13049 & 0.066572 & 0.055875 + & @xmath362 & 0.062407 & 0.065842 & 0.037405 & 0.042978 + + @xmath5 & @xmath358 & 0.013595 & - & 0.0004864 & - + & & & & & + & @xmath359 & 5.3413 & 4.816 & 0.77321 & 0.8116 + & @xmath360 & 0.021 & 0.021903 & 0.033471 & 0.032872 + & @xmath361 & 0.12645 & 0.10498 & 0.052965 & 0.044454 + & @xmath362 & 0.071253 & 0.075175 & 0.042467 & 0.048795 + + + & & & + @xmath210 & error & for @xmath356 & for @xmath357 & for @xmath356 & for @xmath357 + @xmath323 & @xmath358 & 0.67514 & - & 0.21243 & - + & & & & & + & @xmath359 & 1.1052 & 1.1012 & 0.51202 & 0.50951 + & @xmath360 & 0.0062189 & 0.0050111 & 0.074177 & 0.071345 + & @xmath361 & 8.5087 & 8.4753 & 6.4524 & 6.4581 + & @xmath362 & 3.9073 & 3.8963 & 6.1946 & 6.1882 + + @xmath19 & @xmath358 & 0.32181 & - & 0.23381 & - + & & & & & + & @xmath359 & 0.77685 & 0.77406 & 0.35744 & 0.3558 + & @xmath360 & 0.0043659 & 0.00366 & 0.053575 & 0.051756 + & @xmath361 & 5.2257 & 5.2046 & 4.1104 & 4.1143 + & @xmath362 & 3.0451 & 3.0361 & 4.692 & 4.6875 + + @xmath20 & @xmath358 & 0.0046133 & - & 0.0035049 & - + & & & & & + & @xmath359 & 6.6737 & 6.67 & 8.0558 & 8.0599 + & @xmath360 & 7.242 & 7.2428 & 0.37022 & 0.37236 + & @xmath361 & 15.74 & 15.7 & 25.631 & 25.642 + & @xmath362 & 13.328 & 13.306 & 20.092 & 20.077 + + @xmath5 & @xmath358 & 0.0017615 & - & 0.001069 & - + & & & & & + & @xmath359 & 246.55 & 246.41 & 67.29 & 67.32 + & @xmath360 & 9.9591 & 9.9604 & 0.52885 & 0.53228 + & @xmath361 & 13.038 & 13.006 & 19.627 & 19.635 + & @xmath362 & 14.84 & 14.817 & 23.49 & 23.472 + + + & & & + @xmath210 & error & for @xmath356 & for @xmath357 & for @xmath356 & for @xmath357 + @xmath323 & @xmath358 & 0.26501 & - & 1.3628 & - + & & & & & + & @xmath359 & 0.79783 & 0.80083 & 0.90685 & 0.90956 + & @xmath360 & 1.4549 & 1.457 & 1.2481 & 1.2505 + & @xmath361 & 1.3056 & 1.2901 & 2.0582 & 2.0702 + & @xmath362 & 2.3955 & 2.3962 & 0.069749 & 0.072757 + + @xmath19 & @xmath358 & 0.047434 & - & 0.20486 & - + & & & & & + & @xmath359 & 0.5599 & 0.56232 & 0.63362 & 0.63569 + & @xmath360 & 1.0571 & 1.0589 & 0.90402 & 0.90589 + & @xmath361 & 0.82642 & 0.81614 & 1.2756 & 1.2832 + & @xmath362 & 1.836 & 1.8368 & 0.053565 & 0.055977 + + @xmath20 & @xmath358 & 0.003912 & - & 0.0035128 & - + & & & & & + & @xmath359 & 1.218 & 1.2144 & 1.1886 & 1.1919 + & @xmath360 & 4.7735 & 4.775 & 2.1767 & 2.1788 + & @xmath361 & 7.4201 & 7.3951 & 2.2105 & 2.2271 + & @xmath362 & 5.5423 & 5.5439 & 1.9531 & 1.9467 + + @xmath5 & @xmath358 & 0.0013182 & - & 0.00041813 & - + & & & & & + & @xmath359 & 24.251 & 24.157 & 45.122 & 45.27 + & @xmath360 & 6.9147 & 6.9164 & 3.1371 & 3.1398 + & @xmath361 & 5.8866 & 5.8671 & 1.7835 & 1.7967 + & @xmath362 & 6.3419 & 6.3437 & 2.212 & 2.2049 + + + & & & + @xmath210 & error & for @xmath356 & for @xmath357 & for @xmath356 & for @xmath357 + @xmath323 & @xmath358 & 0.093953 & - & 0.20248 & - + & & & & & + & @xmath359 & 0.85205 & 0.86417 & 0.83511 & 0.8363 + & @xmath360 & 1.5557 & 1.5598 & 1.1532 & 1.1601 + & @xmath361 & 0.24334 & 0.33902 & 0.26902 & 0.23261 + & @xmath362 & 1.265 & 1.2405 & 1.2654 & 1.3065 + + @xmath19 & @xmath358 & 0.068203 & - & 0.24097 & - + & & & & & + & @xmath359 & 0.59647 & 0.60537 & 0.585 & 0.58586 + & @xmath360 & 1.1278 & 1.1311 & 0.8372 & 0.84222 + & @xmath361 & 0.15127 & 0.21168 & 0.169 & 0.14618 + & @xmath362 & 0.96943 & 0.95081 & 0.97333 & 1.0049 + + @xmath20 & @xmath358 & 0.0026249 & - & 0.0035651 & - + & & & & & + & @xmath359 & 0.095422 & 0.080499 & 0.18152 & 0.18276 + & @xmath360 & 4.6443 & 4.6454 & 2.382 & 2.3867 + & @xmath361 & 2.3591 & 2.2252 & 2.7907 & 2.8466 + & @xmath362 & 1.6382 & 1.5842 & 1.897 & 1.9858 + + @xmath5 & @xmath358 & 0.00098739 & - & 0.0010871 & - + & & & & & + & @xmath359 & 3.025 & 2.5143 & 4.2062 & 4.237 + & @xmath360 & 6.7404 & 6.7414 & 3.4276 & 3.4343 + & @xmath361 & 1.8975 & 1.7901 & 2.2203 & 2.2648 + & @xmath362 & 1.8705 & 1.8088 & 2.1538 & 2.2546 + + | in this paper we first apply the general analysis described in our first paper to a binary mixture of cyclohexane and @xmath0-hexane .
we use the square gradient model for the continuous description of a non - equilibrium surface and obtain numerical profiles of various thermodynamic quantities in various stationary state conditions . in the second part of this paper we focus on the verification of local equilibrium of the surface as described with excess quantities .
we give a definition of the temperature and chemical potential difference for the surface and verify that these quantities are independent of the choice of the dividing surface .
we verify that the non - equilibrium surface can be described in terms of gibbs excess densities which are in good approximation equal to their equilibrium values at the temperature and chemical potential difference of the surface . |
compact binary star systems are likely to be an important source of gravitational waves for the broadband laser interferometric detectors now under construction @xcite , as they are the best understood of the various types of postulated gravity wave sources in the detectable frequency band and their waves should carry a large amount of information . within our own galaxy , there are three known neutron star binaries whose orbits will decay completely under the influence of gravitational radiation reaction within less than one hubble time , and it is almost certain that there are many more as yet undiscovered .
current estimates of the rate of neutron star / neutron star ( ns / ns ) binary coalescences @xcite based on these ( very few ) known systems project an event rate of three per year within a distance of roughly 200 mpc ; and estimates based on the evolution of progenitor , main - sequence binaries suggest a distance of as small as roughly 70 mpc for three events per year .
these distances correspond to a signal strength which is within the target sensitivities of the ligo and virgo interferometers @xcite .
however , to find the signals within the noisy ligo / virgo data will require a careful filtering of the interferometer outputs . because the predicted signal strengths lie so close to the level of the noise
, it will be necessary to filter the interferometer data streams in order to detect the inspiral events against the background of spurious events generated by random noise .
the gravitational waveform generated by an inspiraling compact binary has been calculated using a combination of post - newtonian and post - minkowskian expansions to post@xmath9-newtonian order by the consortium of blanchet , damour , iyer , will , and wiseman @xcite , and will be calculated to post@xmath10-newtonian order long before the ligo and virgo interferometers come on - line ( c. 2000 ) . because the functional form of the expected signal is so well - known , it is an ideal candidate for matched filtering , a venerable and widely known technique which is laid out in detail elsewhere @xcite and briefly summarized here : the matched filtering strategy is to compute a cross - correlation between the interferometer output and a template waveform , weighted inversely by the noise spectrum of the detector .
the signal - to - noise ratio is defined as the value of the cross - correlation of the template with a particular stretch of data divided by the rms value of the cross - correlation of the template with pure detector noise . if the signal - to - noise ratio exceeds a certain threshold , which is set primarily to control the rate of false alarms due to fluctuations of the noise , a detection is registered .
if the functional form of the template is identical to that of the signal , the mean signal - to - noise ratio in the presence of a signal is the highest possible for any linear data processing technique , which is why matched filtering is also known as optimal filtering . in practice , however , the template waveforms will differ somewhat from the signals .
true gravitational - wave signals from inspiraling binaries will be exact solutions to the einstein equations for two bodies of non - negligible mass , while the templates used to search for these signals will be , at best , finite - order approximations to the exact solutions . also , true signals will be characterized by many parameters ( e.g. the masses of the two objects , their spins , the eccentricity and orientation of the orbit ... ) , some of which might be neglected in construction of the search templates .
thus , the true signals will lie somewhat outside the submanifold formed by the search templates in the full manifold of all possible detector outputs ( see fig .
[ fig : manifold ] ) .
apostolatos @xcite has defined the `` fitting factor '' @xmath11 to quantitatively describe the closeness of the true signals to the template manifold in terms of the reduction of the signal - to - noise ratio due to cross - correlating a signal lying outside the manifold with all the templates lying inside the manifold .
if the fitting factor of a template family is unity , the signal lies in the template manifold .
if the fitting factor is less than unity , the signal lies outside the manifold , and the fitting factor represents the cross - correlation between the signal and the template nearest it in the template manifold .
even if the signal were to lie within the template manifold , it would not in general correspond to any of the actual templates used to search the data .
the parameters describing the search templates ( masses , spins , etc . ) can vary continuously throughout a finite range of values .
the set of templates characterized by the continuously varying parameters is of course infinite , so the interferometer output must be cross - correlated with a finite subset of the templates whose parameter values vary in discrete steps from one template to the next .
this subset ( the `` discrete template family '' ) has measure zero on the manifold of the full set of possible templates ( the `` continuous template family '' ) , so the template which most closely matches a signal will generally lie in between members of the discrete template family ( again , see fig . [
fig : manifold ] ) .
the mismatch between the signal and the nearest of the discrete templates will cause some reduction in the signal - to - noise ratio and therefore in the observed event rate , as some signals which would lie above the threshold if cross - correlated with a perfectly matched filter are driven below the threshold by the mismatch .
thus the spacing between members of the discrete template family must be chosen so as to render acceptable the loss in event rate , without requiring a prohibitive amount of computing power to numerically perform the cross - correlations of the data stream with all of the discrete templates .
the high computational demands of a laser interferometric detector may in fact make it desirable to perform a _
hierarchical search_. in a hierarchical search , each stretch of data is first filtered by a set of templates which rather sparsely populates the manifold , and stretches which fail to exceed a relatively low signal - to - noise threshold are discarded .
the surviving stretches of data are filtered by a larger set of templates which more densely populates the manifold , and are subjected to a higher threshold .
the spareseness of the first - pass template set insures that most of the data need only be filtered by a small number of templates , while the high threshold of the final pass reduces the false alarm rate to an acceptable level .
theoretical foundations for choosing the discrete set of templates from the continuous family were laid by sathyaprakash and dhurandhar for the case of white noise in ref . , and
for ( colored ) power - recycling interferometer noise in ref . .
both papers used a simplified ( so - called `` newtonian '' ) version of the waveform which can be characterized by a single parameter , the binary s `` chirp mass '' @xmath12 .
recently , sathyaprakash @xcite began consideration of an improved , `` post - newtonian '' set of templates characterized by two mass parameters .
he found that , by a judicious choice of the two parameters , the spacing between templates can be made constant in both dimensions of the intrinsic parameter space .
sathyaprakash s parameters also make it obvious ( by producing a very large spacing in one of the dimensions ) that a two - parameter set of templates can be constructed which , if it does not populate the manifold too densely , need not be much more numerous than the one - parameter set of templates used in refs . . in this paper
i shall recast the s - d formalism in geometric language which , i believe , simplifies and clarifies the key ideas .
i shall also generalize the s - d formalism to an arbitrary spectrum of detector noise and to a set of template shapes characterized by more than one parameter .
this is necessary because , as apostolatos @xcite has shown , no one - parameter set of templates can be used to filter a post - newtonian signal without causing an unacceptably large loss of signal - to - noise ratio . in one respect ,
my analysis will be more specialized than that of the s - d formalism .
my geometric analysis requires that the templates of the discrete set be spaced very finely in order that certain analytical approximations may be made , while the numerical methods of sathyaprakash and dhurandhar are valid even for a large spacing between templates ( as would be the case in the early stages of a hierarchical search ) .
the small spacing approximation is justified on the grounds that at some point , even in a hierarchical search , the data must be filtered by many closely spaced templates in order to detect a reasonable fraction ( of order unity ) of the binary inspirals occurring in the universe within range of the ligo / virgo network .
the rest of this paper is organized as follows : in sec .
[ sec : formalism ] , i develop my generalized , geometric variant of the s - d formalism .
i then apply this formalism to the general problem of detection of gravitational waves from inspiraling binaries , and develop general formulas for choosing a discrete template family from a given continuous template family . in sec .
[ sec:1pn ] , i detail an example of the use of my formalism , choosing discrete templates from a continuous template family which describes nonspinning , circularized binaries to post@xmath1-newtonian order in the evolution of the waveform s phase .
i also estimate the computing power required for a single - pass ( non - hierarchical ) search using this discrete template family , and compare to the previous work of sathyaprakash @xcite . finally , in sec .
[ sec : conclusion ] , i summarize my results and suggest future directions for research on the choice of discrete search templates .
in this section , a geometric , multiparameter variant of the s - d formalism is developed .
unless otherwise stated , the following conventions and definitions are assumed : following cutler and flanagan , we define the inner product between two functions of time @xmath13 and @xmath14 ( which may be templates or interferometer output ) as @xmath15 .\end{aligned}\ ] ] here @xmath16 is the fourier transform of @xmath13 , @xmath17 and @xmath18 is the detector s noise spectrum , defined below .
the interferometer output @xmath19 consists of noise @xmath20 plus a signal @xmath21 , where @xmath22 is a dimensionless , time - independent amplitude and @xmath23 is normalized such that @xmath24 .
thus , @xmath22 describes the strength of a signal and @xmath23 describes its shape .
waveform templates are denoted by @xmath25 , where @xmath26 is the vector of `` intrinsic '' or `` dynamical '' parameters characterizing the template shape and @xmath27 is the vector of `` extrinsic '' or `` kinematical '' parameters describing the offsets of the waveform s endpoint .
examples of intrinsic parameters @xmath28 are the masses and spins of the two objects in a compact binary ; examples of extrinsic parameters @xmath29 are the time of a compact binary s final coalescence @xmath30 and the phase of the waveform at coalescence @xmath31 .
templates are assumed to be normalized such that @xmath32 for all @xmath27 and @xmath26 .
expectation values of various quantities over an infinite ensemble of realizations of the noise are denoted by @xmath33 $ ] .
the interferometer s strain spectral noise density @xmath18 is the one - sided spectral density , defined by @xmath34 = \frac{1}{2}\delta ( f_1 - f_2 ) s_h(f_1)\ ] ] for positive frequencies and undefined for negative frequencies .
the noise is assumed to have a gaussian probability distribution .
newton s gravitational constant @xmath35 and the speed of light @xmath36 are set equal to one . in developing our formalism , we begin by defining the signal - to - noise ratio . for any single template @xmath37 of unit norm , the cross - correlation with pure noise @xmath38
is a random variable with mean zero and variance unity ( cf .
ii.b . of ref . , wherein it is shown that @xmath39 = \langle a | b\rangle$ ] ) . the signal - to - noise ratio of a given stretch of data @xmath19 , after filtering by @xmath37 , is defined to be @xmath40 this ratio is the statistic which is compared to a predetermined threshold to decide if a signal is present . if the template @xmath41 is the same as the signal @xmath42 , it optimally filters the signal , and the corresponding ( mean ) optimal signal - to - noise ratio is @xmath43 & = & { \rm e}[\langle n+{\cal a}u | u\rangle ] \nonumber \\ & = & { \cal a}.\end{aligned}\ ] ] if the template @xmath41 used to filter the data is not exactly the same as the signal @xmath42 , the mean signal - to - noise ratio is decreased somewhat from its optimal value : @xmath44 & = & { \rm e}[\langle n+{\cal a}s | u\rangle ] \nonumber \\ & = & { \cal a}\langle s | u\rangle.\end{aligned}\ ] ] the inner product @xmath45 , which is bounded between zero and one , is the fraction of the optimal @xmath46 $ ] retained in the mismatched filtering case , and as such is a logical measure of the effectiveness of the template @xmath41 in searching for the signal shape @xmath42 . now suppose that we search for the signal with a family of templates specified by an extrinsic parameter vector @xmath27 and an intrinsic parameter vector @xmath26 .
let us denote the values of the parameters of the actual templates by @xmath47 .
for example , @xmath48 might be the value of the time @xmath30 of coalescence for the @xmath49th template in the family , and @xmath50 might be the phase of the @xmath49th template waveform at coalescence .
the search entails computing , via fast fourier transforms ( fft s ) , all the inner products @xmath51 for @xmath52 in these numerical computations , the key distinction between the extrinsic parameters @xmath27 and the intrinsic parameters @xmath26 is this : one explores the whole range of values of @xmath27 very quickly , automatically , and efficiently for a fixed value of @xmath26 ; but one must do these explorations separately for each of the @xmath53 . in this sense , dealing with the extrinsic parameters is far easier and more automatic than dealing with the intrinsic ones . as an example ( for further detail
see sec .
16.2.2 of schutz @xcite ) , for a given stretch of data one explores _
all _ values of the time of coalescence @xmath54 of a compact binary simultaneously ( for fixed values of the other template parameters ) via a single fft .
if we write the fourier transform ( for notational simplicity ) as a continuous integral rather than a discrete sum , we get @xmath55 the discrete fft yields the discrete analog of the function of @xmath30 as shown above , an array of numbers containing the values of the fourier transform for all values of @xmath30 . because , for fixed @xmath53 , the extrinsic parameters @xmath27 are dealt with so simply and quickly in the search , throughout this paper we shall focus primarily on a template family s intrinsic parameters @xmath26 , which govern the shape of the template .
correspondingly , we shall adopt the following quantity as our measure of the effectiveness with which a particular template shape i.e .
a particular vector @xmath53 of the intrinsic parameters matches the incoming signal : @xmath56{c}{\textstyle\max}\\{\scriptstyle\mu}\end{array } \langle s | u(\bbox{\mu},\bbox{\lambda}_{(k ) } ) \rangle .\ ] ] here the maximization is over all continuously varying values of the extrinsic parameters .
then the logical measure of the effectiveness of the entire discrete family of templates in searching for the signal shape is @xmath57{c}{\textstyle\max}\\{\scriptstyle k}\end{array } \bigl [ \begin{array}[t]{c}{\textstyle\max}\\{\scriptstyle\mu}\end{array } \langle s | u(\bbox{\mu},\bbox{\lambda}_{(k ) } ) \rangle \bigr ] , \ ] ] which is simply ( [ sort of match ] ) maximized over all the discrete template shapes . in order to focus on the issue of discretization of the template parameters rather than on the inadequacy of the continuous template family , let us assume that the signal shape @xmath42 is identical to some template .
the discussion of the preceding paragraphs suggests that in discussing the discretization of the template parameters we will want to make use of the _ match _ between two templates @xmath58 and @xmath59 which we will define as @xmath60{c}{\textstyle\max}\\{\scriptstyle\mu,\delta\mu}\end{array } \langle u(\bbox{\mu},\bbox{\lambda } ) | u(\bbox{\mu}+\delta\bbox{\mu},\bbox{\lambda}+\delta\bbox{\lambda } ) \rangle .\ ] ] this quantity , which is known in the theory of hypothesis testing as the _ ambiguity function , _ is the fraction of the optimal signal - to - noise ratio obtained by using a template with intrinsic parameters @xmath26 to filter a signal identical in shape to a template with intrinsic parameters @xmath61 . using the match ( [ match definition ] ) it
is possible to quantify our intuitive notion of how `` close '' two template shapes are to each other .
since the match clearly has a maximum value of unity at @xmath62 , we can expand in a power series about @xmath62 to obtain @xmath63 this suggests the definition of a metric @xmath64 so that the _ mismatch _ @xmath65 between two nearby templates is equal to the square of the proper distance between them : @xmath66 having defined a metric on the intrinsic parameter space , we can now use it to calculate the spacing of the discrete template family required to retain a given fraction of the ideal event rate .
schematically , we can think of the templates as forming a lattice in the @xmath67-dimensional intrinsic parameter space whose unit cell is an @xmath67-dimensional hypercube with sides of proper length @xmath68 .
the worst possible case ( lowest @xmath69 $ ] ) occurs if the point @xmath70 describing the signal is exactly in the middle of one of the hypercubes .
if the templates are closely spaced , i.e. @xmath71 , such a signal has a squared proper distance @xmath72 from the templates at the corners of the hypercube .
we define the _ minimal match _
@xmath73 to be the match between the signal and the nearest templates in this worst possible case , i.e. the fraction of the optimal signal - to - noise ratio retained by a discrete template family when the signal falls exactly `` in between '' the nearest templates .
this minimal match is the same quantity that dhurandhar and sathyaprakash in ref .
denote as @xmath74 ; but since it is the central quantity governing template spacing it deserves some recognition in the form of its own name .
our choice of name closely parallels the term `` fitting factor '' @xmath11 , which apostolatos introduced in ref .
@xcite to measure the similarity between actual signals and a continuous template family .
the minimal match , which is chosen by the experimenter based upon what he or she considers to be an acceptable loss of ideal event rate , will determine our choice of spacing of the discrete template parameters and therefore the number of discrete templates in the family .
more specifically , the experimenter will choose some desired value of the minimal match @xmath73 ; and then will achieve this @xmath73 by selecting the templates to reside at the corners of hypercubes with edge @xmath68 given by @xmath75 the number of templates in the resulting discrete template family will be the proper volume of parameter space divided by the proper volume per template @xmath76 , i.e. @xmath77 the formalism above applies to the detection of any set of signals which have a functional form that depends on a set of parameters which varies continuously over some range .
we now develop a more explicit formula for the metric , given an analytical approximation to the ligo noise curve and a particular class of inspiraling binary signals .
we approximate the `` initial '' and `` advanced '' benchmark ligo noise curves by the following analytical fit to fig .
7 of ref .
@xcite : @xmath78\},&f >
f_s\\ \infty,&f < f_s , \end{array } \right.\ ] ] where @xmath79 is the `` knee frequency '' or frequency at which the interferometer is most sensitive ( which is determined by the reflectivities of the mirrors and is set by the experimenters to the frequency where photon shot noise begins to dominate the spectrum ) and @xmath80 is a constant whose value is not important for our purposes .
this spectrum describes photon shot noise in the `` standard recycling '' configuration of the interferometer ( second term ) superposed on thermal noise in the suspension of the test masses ( first term ) , and it approximates seismic noise by setting @xmath81 infinite at frequencies below the `` seismic - cutoff frequency '' @xmath82 . throughout the rest of this paper ,
the `` first ligo noise curve '' will refer to ( [ ligo noise curve ] ) with @xmath82 = 40 hz and @xmath79 = 200 hz , and the `` advanced ligo noise curve '' will refer to ( [ ligo noise curve ] ) with @xmath82 = 10 hz and @xmath79 = 70 hz . these numbers are chosen to closely fit fig .
7 of ref .
@xcite for the first ligo interferometers and for the advanced ligo benchmark . in this paper ,
when various quantities ( such as the number of discrete templates ) are given including a scaling with @xmath79 , this indicates how the quantity changes while @xmath79 is varied but @xmath83 is held fixed . at this point
it is useful to define the moments of the noise curve ( [ ligo noise curve ] ) , following poisson and will , as [ moment definition ] @xmath84 the upper limit of integration @xmath85 denotes the coalescence frequency or high - frequency cutoff of whatever template we are dealing with , which very roughly corresponds to the last stable circular orbit of a test particle in a non - spinning black hole s schwarzschild geometry .
for both first and advanced ligo noise curves , the majority of inspiraling binary search templates will occupy regions of parameter space for which @xmath85 is many times @xmath79 . because we will always be dealing with @xmath86 for @xmath87 , and because the noise term in the denominator of the integrand in eq .
( [ moment definition ] ) rises as @xmath88 for @xmath89 , we can simplify later calculations by approximating @xmath90 in the definition of the moments . to illustrate the metric formalism , we shall use templates based on a somewhat simplified version of the post - newtonian expansion .
since the inner product ( [ inner product definition ] ) has negligible contributions from frequencies at which the integrand oscillates rapidly , it is far more important to get the phase of @xmath91 right than it is to get the amplitude dependence .
therefore , we adopt templates based on the `` restricted '' post - newtonian approximation in which one discards all multipolar components except the quadrupole , but keeps fairly accurate track of the quadrupole component s phase ( for more details see secs . ii.c . and iii.a . of ref
applying the stationary phase approximation to that quadrupolar waveform , we obtain @xmath92,\ ] ] up to a multiplicative constant which is set by the condition @xmath93 @xcite .
the function @xmath94 , describing the phase evolution in ( [ template definition ] ) , is currently known to post@xmath9-newtonian order for the case of two nonspinning point masses in a circular orbit about each other as @xmath95 \end{array}\ ] ] ( cf .
( 3.6 ) of ref . ) . here
the mass parameters have been chosen to be @xmath96 , the total mass of the system , and @xmath97 , the ratio of the reduced mass to the total mass .
the actual amplitude @xmath22 of a waveform is proportional to @xmath98 , where @xmath99 is the distance to the source ; and this lets us find the relation between minimal match and event rate which we will need in order to wisely choose the minimal match . assuming that compact binaries are uniformly distributed throughout space on large distance scales
, this means that the rate of events with a given set of intrinsic parameters and with an amplitude greater than @xmath22 is proportional to @xmath100 .
setting a signal - to - noise threshold @xmath101 is equivalent to setting a maximum distance @xmath102 to which sources with a given set of intrinsic parameters can be detected .
thus , if we could search for signals with the entire continuous template family , we would expect the observed event rate to scale as @xmath103 .
this ideal event rate is an upper limit on what we can expect with a real , discrete template family .
we can obtain a lower bound on the observed event rate by considering what happens if all signals conspire to have parameters lying exactly in between the nearest search templates . in this case , all events will have reduced signal - to - noise ratios of @xmath73 times the optimal signal - to - noise ratio @xmath22 . this is naively equivalent to optimally filtering with a threshold of @xmath104 , so a pessimistic guess is @xmath105 for a fixed rate of false alarms . in real life
, @xmath101 is affected by the total number of discrete templates and by the minimal match of the discrete template family .
this can be seen by the fact that the signal - to - noise ratio , @xmath106{c}{\textstyle\max}\\{\scriptstyle k}\end{array } \bigl [ \begin{array}[t]{c}{\textstyle\max}\\{\scriptstyle\mu}\end{array } \langle o | u(\bbox{\mu},\bbox{\lambda}_{(k ) } ) \rangle \bigr ] , \ ] ] is the maximum of a number of random variables .
the covariance matrix of these variables will be determined by the minimal match , and will itself determine the probability distribution of @xmath107 ( in the absence of a signal ) which is used to set the threshold @xmath101 in order to keep the false alarm rate below a certain level . however , since these effects are fairly small at high signal - to - noise ratios ( such as those to be used by ligo ) and the issue of choosing thresholds is a problem worthy of its own paper @xcite , we will use ( [ er vs mm ] ) for the rest of this paper .
for this two - parameter template family , the formula for the match ( [ match definition ] ) can be simplified somewhat by explicitly performing the maximization over the extrinsic parameters @xmath27 and @xmath108 .
since the integrand in the inner product ( [ inner product definition ] ) depends on @xmath27 and @xmath108 as @xmath109 $ ] , there is no dependence on @xmath27 but only on @xmath108 .
maximizing over @xmath110 is easy : instead of taking the real part of the integral in the inner product ( [ inner product definition ] ) , we take the absolute value . to maximize over @xmath111 we go back a step .
let us define @xmath112 , and consider the @xmath113-dimensional space formed by @xmath114 and @xmath115 .
we expand the inner product between adjacent templates to quadratic order in @xmath116 to get a preliminary metric @xmath117 , where the greek indices range from @xmath118 to @xmath67 ( latin indices range from @xmath119 to @xmath67 ) : @xmath120\right| } { \int_0^{\infty}df\frac{\displaystyle f^{-7/3}}{\displaystyle s_h(f)}}\right\ } \right ] { \atop\delta\lambda^\alpha = 0}.\ ] ] here @xmath121 @xcite .
we define the moment functional @xmath122 such that , for a function @xmath123 , @xmath124 = \frac{1}{i(7)}\int_{f_s / f_0}^{f_c / f_0 } dx\frac{x^{-7/3}}{s_h(x~f_0)}a(x),\ ] ] and thus @xmath125 = \sum_n a_n j(7 - 3n).\ ] ] we also define the quantities @xmath126 such that @xmath127 where the derivative is evaluated at @xmath128 and @xmath129 is the part of @xmath94 in eq .
( [ template definition ] ) that is frequency dependent ( any non - frequency - dependent , additive parts of @xmath94 are removed when we take the absolute value in the maximized inner product ) .
evaluation of the derivative in eq .
( [ premetric - def ] ) then shows that , in the limit @xmath130 , @xmath131 - { \cal j}[\psi_\alpha]{\cal j}[\psi_\beta]\right).\ ] ] finally , we minimize @xmath132 with respect to @xmath111 ( i.e. , we project @xmath117 onto the subspace orthogonal to @xmath30 ) and thereby obtain the following expression for the metric of our continuous template family : @xmath133 by taking the square root of the determinant of this metric and plugging it into eq .
( [ cal n definition ] ) , we can compute the number of templates @xmath0 that we need in our discrete family as a function of our desired minimal match @xmath73 , or equivalently of the loss of ideal event rate .
although the phase of the inspiraling binary signal has recently been calculated to post@xmath9-newtonian order @xcite , it is useful to calculate the number of templates that would be required in a universe where the waveforms evolve only to post@xmath1-newtonian order and all binaries are composed of nonspinning objects in circular orbits .
there are several reasons for this exercise . 1 .
apostolatos @xcite has shown that amplitude modulation of the waveform due to spin effects is important in an inspiraling binary search only for a few extremal combinations of parameters , and also that ( at higher post - newtonian order ) templates without spin - related phase modulation can match phase modulated signals almost as well as can templates that include spin parameters .
therefore the bulk of the final set of templates actually used when the detectors come on - line will not need to include the extra spin parameters , and we may ignore them in this preliminary work .
we assume circular orbits because gravitational radiation reaction circularizes most eccentric orbits on a timescale much less than the lifetime of a compact binary @xcite .
the phase of the templates is truncated at post@xmath1-newtonian order for simplicity .
although apostolatos has demonstrated in ref .
@xcite that post@xmath1-newtonian templates will not have a large enough fitting factor to be useful , consideration of such a set is a first step toward obtaining an adequate set of templates and it is a particularly important step since the metric coefficients will turn out to be constant over the template manifold . having chosen as the continuous template family the set of post@xmath1-newtonian , circular , spinless binary waveforms , we must now choose the discrete templates from within this continuous family .
the first step is to calculate the coefficients of the metric on the two - dimensional dynamical parameter space .
it is convenient to change the mass parameterization from the variables @xmath134 to the sathyaprakash variables @xcite @xmath135 note that @xmath136 and @xmath137 are simply the newtonian and post@xmath1-newtonian contributions to the time it takes for the carrier gravitational wave frequency to evolve from @xmath79 to infinity .
the advantage of these variables is that the metric coefficients in @xmath138 coordinates are constant ( in the limit @xmath139 ) for all templates .
this is because the phase of the waveform @xmath91 is linear in the sathyaprakash variables , and so the integral in the definition of the match ( [ match definition ] ) depends only on the displacement @xmath140 between the templates , not on the location @xmath138 of the templates in the dynamical parameter space .
the dynamical parameter - dependent part of the templates phase is given by [ eq .
( [ 2pn psi ] ) truncated to first post - newtonian order and reexpressed in terms of @xmath138 using eqs .
( [ def : tau1 ] ) and ( [ def : tau2 ] ) ] @xmath141 and it is easy to read off @xmath142 and @xmath143 [ eq .
( [ psi - def ] ) ] as the coefficients of @xmath136 and @xmath137 @xcite . by inserting these @xmath144 into eq .
( [ series functional ] ) , the relevant moment functionals can be expressed in terms of the moments of the noise : @xmath145 & = & 2\pi f_0~j(4 ) , \\ { \cal j}[\psi_1 ] & = & 2\pi f_0~\frac{3}{5}j(12 ) , \\ { \cal j}[\psi_2 ] & = & 2\pi f_0~j(10 ) , \\ { \cal j}[\psi_0 ^ 2 ] & = & ( 2\pi f_0)^2~j(1 ) ,
\\ { \cal j}[\psi_0\psi_1 ] & = & ( 2\pi f_0)^2~\frac{3}{5}j(9 ) , \\ { \cal j}[\psi_0\psi_2 ] & = & ( 2\pi f_0)^2~j(7 ) , \\ { \cal j}[\psi_1 ^ 2 ] & = & ( 2\pi f_0)^2~\frac{9}{25}j(17 ) , \\ { \cal j}[\psi_1\psi_2 ] & = & ( 2\pi f_0)^2~\frac{3}{5}j(15 ) , \\
{ \cal j}[\psi_2 ^ 2 ] & = & ( 2\pi f_0)^2~j(13 ) .
\end{array}\ ] ] we can compute the needed moments of the noise by numerically evaluating the integrals ( [ moment definition ] ) . by setting the upper limit of integration to infinity , i.e. by approximating @xmath146 as infinite for all templates under consideration , we find that the moments have the constant values given in table [ tab : j s ] ; and therefore the moment functionals ( [ 1pn functionals ] ) have the constant values given in table [ tab : cal j s ] . inserting these values into eqs .
( [ premetric ] ) and ( [ metric conversion ] ) yields , for the coordinates @xmath147 , the 3-metric and 2-metric @xmath148 @xmath149 for the first ligo noise curve , and @xmath150 @xmath151 for the advanced ligo noise curve ( where the dots denote terms obtained by symmetry ) .
we shall also estimate the errors in the metric coefficients due to the approximation @xmath152 : the moment integrals defined in eq .
( [ moment definition ] ) can be rewritten as @xmath153 where the first integral is the expression used in the above metric coefficients and the second is the correction due to finite @xmath146 .
the second integral can be expanded to lowest order in @xmath154 as @xmath155 and from this the errors in the moments ( and therefore in the metric coefficients ) due to approximating @xmath85 as infinite are estimated to be less than or of order ten percent for the first ligo interferometers and one percent for the advanced ligo interferometers over most of the relevant volume of parameter space . since the two - parameter , post@xmath1-newtonian continuous template family is known to be inadequate for the task of searching for real binaries , these errors are small enough to justify our use of the @xmath156 approximation in this exploratory analysis .
since the metric coefficients are constant in this analysis , the formula for the required number of templates [ eq . ( [ cal n definition ] ) ] reduces to @xmath157 the square root of the determinant of the metric is given by @xmath158 for the advanced ligo noise curve and by @xmath159 for the initial ligo noise curve , so once we have decided on the range of parameters we deem astrophysically reasonable we will have a formula for @xmath0 as a function of @xmath73 .
the most straightforward belief to cherish about neutron stars is that they all come with masses greater than a certain minimum @xmath3 , which might be set to @xmath160 ( based on the minimum mass that any neutron star can have ) or @xmath161 ( based on the observed masses of neutron stars in binary pulsar systems @xcite ) . in terms of the variables @xmath134 the constraint @xmath162 and @xmath163 is easily expressed as @xmath164 but in terms of the sathyaprakash variables [ eqs .
( [ def : tau1 ] ) and ( [ def : tau2 ] ) ] the expression becomes rather unwieldy to write down .
however , see fig .
[ fig : volume ] for a plot of the allowed region in @xmath138 coordinates . for this reason we have found it convenient to use a monte carlo integration routine @xcite to evaluate the coordinate volume integral @xmath165 .
the monte carlo approach becomes especially attractive when evaluating the proper volume integral @xmath166 for cases where the integrand is allowed to vary and in fact may itself have to be evaluated numerically , as will be the case for a post@xmath9-newtonian set of templates .
the integral has numerical values of @xmath167 and @xmath168 seconds@xmath9 for initial and advanced ligo interferometer parameters , respectively , assuming a @xmath3 of @xmath160 and arbitrarily large @xmath169 .
the integral can be shown ( numerically ) to scale roughly as @xmath170 ( independent of @xmath79 ) and as @xmath171 for @xmath3 ranging from 0.2 to 1.0 solar masses ( the dependence on @xmath169 is negligible for any value greater than a few solar masses ) . inserting the above numbers into eq .
( [ flat cal n ] ) , we find that @xmath172 for the first ligo noise curve and @xmath173 for the advanced ligo noise curve . the fiducial value of @xmath73 has been chosen as 0.97 to correspond to an event rate of roughly 90 percent of the ideal event rate [ cf .
( [ er vs mm ] ) ] . in terms of the template - spacing - induced fractional loss @xmath2 of event rate ,
the number of templates required is @xmath174 for the first ligo noise curve and @xmath175 for the advanced ligo noise curve . with the aid of the metric coefficients given in eqs .
( [ 1st 1pn metric ] ) and ( [ adv 1pn metric ] ) , it is a simple task to select the locations of the templates and the spacing between them .
because the metric coefficients form a constant @xmath176 matrix , we can easily find the eigenvectors @xmath177 and @xmath178 of @xmath179 and use them as axes to lay out a grid of templates .
the numerical values are @xmath180 for the first ligo noise curve and @xmath181 for the advanced ligo noise curve .
the infinitesimal proper distance is given in terms of the eigen - coordinates as @xmath182 , where @xmath183 and @xmath184 are the eigenvalues of the metric .
therefore we simply use eq .
( [ mm vs dl ] ) to obtain the template spacings @xmath185 we find that the eigenvalues of the metrics ( [ 1st 1pn metric ] ) and ( [ adv 1pn metric ] ) are @xmath186 times 0.721 and 0.00427 ( first ligo ) , and @xmath186 times 1.25 and 0.00984 ( advanced ligo )
. therefore the template spacings are given by @xmath187 for the first ligo noise curve and by @xmath188 for the advanced ligo noise curve .
figure [ fig : contours ] shows the locations of some possible templates superposed on a contour plot of the match with the template in the center of the graph .
drawing on the previous work of schutz @xcite concerning the mechanics of fast - fourier - transforming the data , we can estimate the cpu power required to process the interferometer output on - line through a single - pass ( non - hierarchical ) search involving @xmath0 templates .
although the data will be sampled at a rather high rate ( tens of khz ) , frequencies above some upper limit @xmath189 can be thrown away ( in fourier transforming the data ) with only negligible effects on the signal - to - noise ratio .
this lowers the effective frequency of sampling to @xmath190 ( the factor of two is needed so that the nyquist frequency is @xmath191 ) , and thereby considerably reduces the task of performing the inner - product integrals . if the length of the array of numbers required to store a template is @xmath192 and that required to store a given stretch of data is @xmath193 , the number of floating point operations required to process that data stretch through @xmath0 filters is @xmath194 [ cf .
( 16.37 ) of schutz @xcite , with the fractional overlap between data segments @xmath195 chosen as roughly @xmath196 .
actually , @xmath192 varies from filter to filter , but most of the search templates occupy regions of parameter space where the mass is very low and thus the storage size of the template , @xmath197 is very large @xcite .
the longest filter is the one computed for two stars of mass @xmath3 , so by inserting @xmath198 and @xmath199 into eq .
( [ def : tau1 ] ) and combining with eq .
( [ storage size ] ) , we find that we can make a somewhat pessimistic estimate of the computational cost by using @xmath200 the required cpu power @xmath6 for an on - line search is obtained by dividing eq .
( [ n operations ] ) by the total duration of the data set , @xmath201 to find that @xmath202 combining eq .
( [ cpu power ] ) with eqs .
( [ 1st cal n ] ) and ( [ adv cal n ] ) gives us @xmath203 for an on - line search by the first ligo interferometers and @xmath204 for the advanced ligo interferometers .
although the estimates in the paragraph above are not to be believed beyond a factor of order unity , the magnitude of the numbers shows that a hierarchical search strategy may be desirable to keep the computing power requirements at a reasonable level for non - supercomputing facilities .
that is , the data would first be filtered through a more widely spaced ( low minimal match ) set of templates with a relatively low signal - to - noise threshold , and only the segments which exceed this preliminary threshold would be analyzed with the finely spaced ( high minimal match ) templates .
the metric - based formalism of this paper only holds for the finely spaced set of templates used in the final stage of the hierarchical search ; the template spacing used in the earlier stages of the search will have to be chosen using more complex methods such as those of sathyaprakash and dhurandhar .
the only previous analysis of the problem of choosing the discrete search templates from the two - parameter , restricted post@xmath1-newtonian continuous template family is that of sathyaprakash @xcite , in which he found that the entire volume of parameter space corresponding to @xmath205 could be covered by a set of templates which vary only in @xmath206thereby reducing the effective dimensionality of the mass parameter space to one .
this implied a value of @xmath0 similar to that obtained in the one - parameter ( newtonian template ) analysis of dhurandhar and sathyaprakash in ref . .
it is not possible to fairly compare my value for @xmath0 to the values given by dhurandhar and sathyaprakash in table ii of ref . due to our differing assumptions concerning the sources and the desirable level of the minimal match
therefore i will compare the assumptions .
dhurandhar and sathyaprakash typically consider a minimal match of 0.8 or 0.9 rather than 0.97 , this would lead to a loss of thirty to fifty percent of the ideal event rate [ cf .
( [ er vs mm ] ) ] .
if the current `` best estimates '' of inspiraling binary event rates @xcite are correct , the ideal event rate for ligo and virgo will not be more than about one hundred per year even when operating at the `` advanced interferometer '' noise levels , and the loss of up to half of these events would be unacceptable . from eqs .
( [ 1st cal n ] ) and ( [ adv cal n ] ) it can be seen that the dependence of @xmath0 on @xmath3 is the most important factor influencing the number of templates . the two - parameter analysis of sathyaprakash @xcite uses a value for @xmath3 of @xmath207 , which is based on the statistics of ( electromagnetically- ) known binary pulsars .
however , because there is no known , firm physical mechanism that prevents neutron stars from forming with masses between 0.2 and 1 @xmath208 , ligo and virgo should use a discrete template family with @xmath209 .
after all , laser interferometer gravitational wave detectors are expected to bring us information about astronomical objects as yet unknown . during the final stages of completion of this manuscript ,
a new preprint by balasubramanian , sathyaprakash , and dhurandhar appeared in the xxx.lanl.gov archive @xcite , applying differential geometry to the problem of detecting compact binary inspiral events and extracting source parameters from them .
the preprint applies the tools of differential geometry primarily to the problem of parameter measurement rather than that of signal detection , and so does not develop the geometrical formalism as far as is done in sec . [
sec : formalism ] of this paper .
the metric constructed in ref .
@xcite is identical to the information matrix which was suggested for use in the construction of a closely - spaced discrete template family in the authors previous work .
while this is quite useful for parameter measurement , it neglects maximization over kinematical parameters and thus is not very useful for the construction of search templates .
also , the assumptions about @xmath73 and @xmath3 are no different from those made in previous work up to and including ref .
@xcite , and so the result for @xmath0 is no different . the main difference between the results of ref .
@xcite and previous analyses by the same authors and therefore the most important part of the preprint as far as the detection problem is concerned is the introduction of the possibility of choosing search templates to lie outside the manifold of the continuous template family . using an _ ad hoc _ example ,
the authors show that such a placement of templates can result in a spacing roughly double that between discrete templates chosen from the manifold formed by the continuous template family .
my analysis in this paper does not consider this possibility , but the formalism of sec .
[ sec : formalism ] can easily be extended to investigate this problem in the future .
this paper has presented a method for semi - analytically calculating the number of templates required to detect gravity waves from inspiraling binaries with ligo as a function of the fraction of event rate lost due to the discrete spacing of the templates in the binary parameter space .
this method , based on differential geometry , emphasizes that ultimately a finer template spacing is required than has previously been taken as typical in the literature , in order to retain a reasonable fraction of the event rate .
this paper details the first calculation of this kind that uses post - newtonian templates and a noise curve which takes into account the coloration of noise in the detector due to both standard recycling photon shot noise and thermal noise in the suspension of the test masses .
the result is that it is possible to search the data for binaries containing objects more massive than @xmath160 thoroughly enough to lose only @xmath210 10 percent of the ideal event rate without requiring a quantum leap in computing technology .
the computational cost of such a search , conducted on - line using a single pass through the data , is roughly 20 gflops for the first ligo interferometers ( ca .
2000 ) and 270 gflops for the advanced ligo interferometers ( some years later ) .
this is feasible ( or very nearly feasible ) even for a present - day supercomputing facility , but a hierarchical search strategy ( using as its first stage a widely - spaced set of templates similar to that analyzed by sathyaprakash @xcite ) may be desirable to reduce the cost .
a thorough investigation of hierarchical search strategies is in order : how should the threshold and the minimal match of the first stage be set in order to minimize the cpu power required while keeping the false alarm and false dismissal rates at acceptable levels ?
how would non - gaussian noise statistics affect the first stage threshold and minimal match ?
would a hierarchical search benefit by using more than two stages ?
how is the threshold affected by the minimal match when the approximation of high signal - to - noise ratio can no longer be made ?
the formalism of this paper should be applied to choose discrete templates from a better continuous template family than the one considered here .
the best two - parameter templates will be based on the highest post - newtonian order computations that have been performed for circularized , spinless binaries , augmented perhaps by terms of still higher order from the theory of perturbations of schwarzschild or kerr spacetime .
i plan to soon apply my geometric formalism to the post@xmath211-newtonian templates which are currently the best available . the areas of parameter space where spins can not be neglected ( noted by apostolatos in ref .
@xcite ) must also be investigated , and the inclusion of an orbital eccentricity parameter should be considered .
there are several more issues which i plan to address using my formalism or some extension of it .
an analysis needs to be made for the case when the signal is not identical to some member of the continuous template family ( i.e. the fitting factor is not equal to one ) ; and the result of such an analysis should be used to set definite goals for both the fitting factor and the minimal match in terms of event rate .
the effect of non - quadrupolar harmonics of the gravitational wave on the construction of search templates should be considered .
these harmonics have been ignored in all previous analyses of detection and even of parameter measurement , but they may have a noticeable effect when a very high minimal match is desired . finally , a systematic investigation of the optimal choice of search templates outside the continuous template family is in order .
this problem has been addressed in a preliminary way in ref .
@xcite , but is deserving of further scrutiny . my thanks to theocharis apostolatos and eanna flanagan for helping me get started .
i am most indebted to kip thorne for his guidance and his patience in reviewing the manuscript .
this work was supported in part by my nsf graduate fellowship and in part by nsf grant phy-9424337 .
a. abramovici , w. e. althouse , r. w. p. drever , y. grsel , s. kawamura , f. j. raab , d. shoemaker , l. sievers , r. e. spero , k. s. thorne , r. e. vogt , r. weiss , s. e. whitcomb , and m. e. zucker , science * 256 * , 325 ( 1992 ) . c. bradaschia , e. calloni , m. cobal , r. del fasbro , a. di virgilio , a. giazotto , l. e. holloway , h. kautzky , b. michelozzi , v. montelatici , d. pascuello , and w. velloso , in _ gravitation 1990 _ , proceedings of the banff summer institute , ed .
r. mann and p. wesson ( world scientific , singapore , 1991 ) . see , e.g. , n. wiener , _ the extrapolation , interpolation and smoothing of stationary time series with engineering applications _
( wiley , new york , 1949 ) or l. a. wainstein and v. d. zubakov , _ extraction of signals from noise _
( prentice - hall , london , 1962 ) .
note that the @xmath31 used here is the coalescence phase of the _ waveform_. in the restricted post - newtonian approximation , this is twice the coalescence phase of the _ orbit _ @xmath212 , which is used by cutler and flanagan in ref . .
this @xmath113-dimensional metric @xmath117 is related to the fisher information matrix , which was used by dhurandhar and sathyaprakash in the appendix of ref . to compute the template spacing analytically in the small - spacing limit . however , in taking the absolute value to maximize over @xmath110 , we have lost some dependence on the dynamical ( intrinsic ) template parameters , so our @xmath213-metric is not even the projection of the information matrix onto the subspace perpendicular to @xmath31 .
this lets us tolerate a greater mismatch between the templates dynamical parameters than was possible before maximization , thereby reducing the total number of templates required .
the @xmath144 given here are different from those given in eq .
( 13a-3 ) of ref .
@xcite because sathyaprakash does not use @xmath30 and @xmath31 as kinematical parameters , but rather the time and phase at which the quadrupole part of the waveform reaches a frequency of @xmath79 .
numerical integrations were performed with the aid of mathematica [ s. wolfram , _ mathematica : a system for doing mathematics by computer _ ( addison - wesley , redwood city , california , 1988 ) ] and with a monte carlo code based on numerical recipes [ w. h. press , s. a. teukolsky , w. t. vetterling , and b. p. flannery , _ numerical recipes : the art of scientific computing _
( cambridge university press , cambridge , england , 1992 ) ] .
the storage size of the template is equal to the effective sampling frequency @xmath190 times the duration of the template waveform . in the restricted post - newtonian approximation ,
the duration of a binary chirp waveform template is equal to the time it takes for the gravitational wave frequency to sweep from the seismic frequency @xmath82 up to infinity . since @xmath214 is the dominant ( newtonian ) contribution to the time it takes the gravitational wave frequency to sweep from @xmath79 up to infinity
, the duration of a waveform template is roughly @xmath215 .
thus we obtain eq .
( [ storage size ] ) .
r. balasubramanian , b. s. sathyaprakash , and s. v. dhurandhar , _ gravitational waves from coalescing binaries : detection strategies and monte carlo estimation of parameters _ , submitted to phys .
d ( los alamos xxx.lanl.gov archive preprint gr - qc/9508011 ) . | gravitational waves from inspiraling , compact binaries will be searched for in the output of the ligo / virgo interferometric network by the method of `` matched filtering''i.e . , by correlating the noisy output of each interferometer with a set of theoretical waveform templates .
these search templates will be a discrete subset of a continuous , multiparameter family , each of which approximates a possible signal .
the search might be performed _ hierarchically _ , with a first pass through the data using a low threshold and a coarsely - spaced , few - parameter template set , followed by a second pass on threshold - exceeding data segments , with a higher threshold and a more finely spaced template set that might have a larger number of parameters .
alternatively , the search might involve a single pass through the data using the larger threshold and finer template set .
this paper extends and generalizes the sathyaprakash - dhurandhar ( s - d ) formalism for choosing the discrete , finely - spaced template set used in the final ( or sole ) pass through the data , based on the analysis of a single interferometer .
the s - d formalism is rephrased in geometric language by introducing a metric on the continuous template space from which the discrete template set is drawn .
this template metric is used to compute the loss of signal - to - noise ratio and reduction of event rate which result from the coarseness of the template grid .
correspondingly , the template spacing and total number @xmath0 of templates are expressed , via the metric , as functions of the reduction in event rate .
the theory is developed for a template family of arbitrary dimensionality ( whereas the original s - d formalism was restricted to a single nontrivial dimension ) .
the theory is then applied to a simple post@xmath1-newtonian template family with two nontrivial dimensions . for this family
, the number of templates @xmath0 in the finely - spaced grid is related to the spacing - induced fractional loss @xmath2 of event rate and to the minimum mass @xmath3 of the least massive star in the binaries for which one searches by @xmath4 for the first ligo interferometers and by @xmath5 for advanced ligo interferometers .
this is several orders of magnitude greater than one might have expected based on sathyaprakash s discovery of a near degeneracy in the parameter space , the discrepancy being due to this paper s lower choice of @xmath3 and more stringent choice of @xmath2 .
the computational power @xmath6 required to process the steady stream of incoming data from a single interferometer through the closely - spaced set of templates is given in floating - point operations per second by @xmath7 for the first ligo interferometers and by @xmath8 for advanced ligo interferometers .
this will be within the capabilities of ligo - era computers , but a hierarchical search may still be desirable to reduce the required computing power . |
the standard model of electroweak interactions predicts the existence of a phase transition inbetween a low temperature symmetry broken and a high temperature symmetric phase @xcite .
its thermodynamic properties lead to cosmological consequences .
one might hope that the baryon asymmetry can be generated at the electroweak phase transition , if the transition is of strong first order . for values of the zero temperature higgs - mass @xmath3
gev the phase transition is of first order @xcite .
as the higgs - mass is increased further , the thermodynamic singularity at the electroweak phase transition weakens .
it is even possible , that for large values of the higgs - mass the phase transition loses its nonanalytical structure and is described by a crossover @xcite . in the strict sense the electroweak phase transition would then cease to exist .
the exact determination of the critical higgs - mass value @xmath4 , at which the first order electroweak phase transition changes from first order to a crossover is important in view of its implications for the standard model .
another aspect of the electroweak theory is the non - vanishing magnetic screening mass .
it controls the infrared behavior of the theory at high temperatures and influences the nature of the phase transition itself .
this work is a continuation of earlier studies and for details on the formulation of the theory we refer to @xcite .
here we recall the simulated @xmath5 action functional : @xmath6 the relationship of the dimensionless lattice couplings @xmath7 to the couplings of the @xmath8 @xmath9 gauge - higgs system can be found in @xcite .
our simulation is performed at @xmath10 and at @xmath11 values @xmath12 and @xmath13 .
the hopping parameter is varied across the phase transition .
these @xmath14-values correspond to higgs - mass values of @xmath15 and @xmath16 gev , if the 1-loop parameter mapping of @xcite is used .
for possible correction terms inducing small corrections to the here cited higgs - mass values we refer to a forthcoming publication .
to illustrate the critical behavior as a function of @xmath14 , or the higgs - mass , we display in figure [ cv_max ] the maxima of the @xmath17-susceptibility , which is given by @xmath18 3.0 in -0.5 in cv_max cv_max.eps maxima of the @xmath17-susceptibility .
we investigate lattice sizes ranging from @xmath19 up to @xmath20 .
the labeling of the data in figure [ cv_max ] corresponds to the one used in figure [ l_im ] .
the data at @xmath21 are consistent with a first order transition , the dotted line with slope @xmath22 in the figure . at @xmath23 and @xmath13 a crossover
is observed , while at @xmath24 the behavior is almost critical , the solid scaling curve in figure [ chi ] .
the determination of the critical higgs - mass value relies on the analysis of fisher or lee - yang zeroes @xcite in the crossover - region of the theory .
the partition function @xmath25 is analytically continued into the complex plane as a function of the complex hopping parameter @xmath26 . denoting with @xmath27 the lowest zero of @xmath25 , i.e. the zero in @xmath26 with smallest length , we expect in the vicinity of the critical end - point the scaling law @xmath28 such a scaling behavior can also be observed for lee - yang zeroes in the high temperature phase of the ising model .
our strategy to localize the end - point then is to determine the value of @xmath14 , at which the regular contribution @xmath29 to the scaling law vanishes : @xmath30 . in figure [ l_im ]
we display @xmath31 vs. @xmath32 .
the solid curves in figure [ l_im ] correspond to fits with the scaling law .
3.0 in -0.5 in l_im l_im.eps imaginary parts of the lowest zeroes of the partition function . in figure [ fitconst ] the constant @xmath33 is displayed as function of @xmath14 .
the data are consistent with a linear dependence on @xmath14 and the fit results into @xmath34 , corresponding to a critical higgs - mass value of approximately @xmath35 gev .
3.0 in -0.5 in fitconst fitconst.eps the determination of @xmath36 .
the w - boson mass is determined from the w - boson propagator in landau - gauge , @xmath37 . for
a detailed discussion of the implementation of the landau - gauge on a lattice see @xcite and references therein .
we investigate the w - boson propagator at @xmath24 , which is very close to the critical higgs - mass .
simulations have been performed on a @xmath38 lattice .
the propagators have been analyzed in the same way as discussed in @xcite .
our results are shown in figure [ mass_w ] . 3.0 in -0.6 in mass_w mass_w.eps w - boson screening masses , calculated on a @xmath38 lattice at @xmath24 .
the full curve describes the fit to the data .
the dashed curve represents the result obtained from gap equations . in the high temperature phase
the w - boson propagator stays constant .
a fit to the data for @xmath39 , the full triangles in the figure , yields @xmath40 .
it is worthwhile to note , that in pure su(2 ) gauge theory the magnetic screening mass has a value @xmath41 , the horizontal full and dotted lines in figure [ mass_w ] . in the symmetry
broken phase the mass increases rapidly . in this region
the data are very well described by the ansatz @xmath42 the fit to the full circles in the figure results into a value for the exponent @xmath43 of @xmath44 .
we are now able to compare our results with predictions based on gap equations @xcite . using similar parameters in the gap equations as in our study
we observe a qualitative agreement , the dashed curve in the figure .
we have exploited the scaling behavior of partition function zeroes in the vicinity of the critical end - point in order to determine the critical higgs - mass @xmath4 .
the electroweak phase transition looses its first order character at a higgs - mass value of about @xmath35 gev .
close to the critical higgs - mass we have measured the w - boson propagator in landau gauge .
w - boson screening masses remain constant in the high temperature symmetric phase and increase in the low temperature higgs phase .
the agreement of these data with predictions based on gap equations is of qualitative nature . | we study the effective 3-d su(2 ) gauge - higgs model at finite temperature for higgs - masses in the range from @xmath0 gev up to @xmath1 gev .
the first order electroweak phase transition weakens with increasing higgs - mass and terminates at a critical end - point . for higgs - mass values larger than about @xmath2 gev
the thermodynamic signature of the transition is described by a crossover .
close to this higgs - mass value we investigate the vector boson propagator in landau gauge .
the calculated w - boson screening masses are compared with predictions based on gap equations .
epsf.sty # 1 # 2 # 3 # 4 # 5 |
ballistic transport of particles across billiards is a field of major importance due to its fundamental properties as well as physical applications ( see for example the reviews @xcite ) . in such systems ,
a two - dimensional cavity is defined by a steplike single - particle potential where confined particles can propagate freely between bounces at the billiard walls . for open systems
the possibility of particles being injected and escaping through holes in the boundary is also allowed . as an example , we consider the open geometry of the extensively studied sinai billiard shown in fig . [ fig : fig1 ] .
experimental realizations are based on exploiting the analogy between quantum and wave mechanics in either microwave and acoustic cavities or vibrating plates @xcite , and on structured two - dimensional electron gases in artificially tailored semiconductor heterostructures @xcite . in the latter case , the particles are also charge carriers making these nanostructures relevant to applied electronics .
the open geometry of the sinai billiard considered in this study.,width=302 ] focussing the attention on the electronic analogues , more recently the possibility to couple a superconductor to a ballistic quantum dot has been considered both theoretically @xcite and experimentally @xcite , so that some part of the billiard boundary exerts the additional property of andreev reflection @xcite . during this process particles with energies much smaller than the superconducting gap @xmath1
are coherently scattered from the superconducting interface as fermi sea holes back to the normal conducting system ( and vice versa ) .
classically , andreev reflection manifests itself by retroreflection , i.e. , all velocity components are inverted , compared to the specular reflection where only the boundary normal component of the velocity is inverted .
thus , andreev reflected particles ( holes ) retrace their trajectories as holes ( particles ) .
if , however , a perpendicular magnetic field is applied in addition , such retracing no longer occurs due to the inversion of both the charge and the effective mass of the quasiparticle resulting in opposite bending .
typical trajectories are illustrated in fig .
[ fig : fig2 ] . here , we investigate the interplay between trajectory bending and andreev reflection and
demonstrate how such effects influence the overall ( magneto)transport properties of andreev billiards when compared to their normal counterparts .
typical specular ( sr ) and andreev reflection ( ar ) at the circular central `` antidot '' of fig .
[ fig : fig1 ] .
a magnetic field is applied as indicated . ]
a unique feature of this class of ( quantum ) mechanical systems is their suitability for studying the quantum - to - classical correspondence . in particular
, much effort has been devoted in revealing the quantum fingerprints of the classical dynamics which may be parametrically tuned from regular to chaotic via , e.g. , changes in the billiard - shape .
a range of theoretical tools has been used , spanning the usual analysis of classical trajectories and the semiclassical approximation to the models of random matrix theory and fully quantum mechanical calculations .
the main signatures of classical integrability ( or lack of it ) on the statistics of energy levels and properties of the transport coefficients for closed and open systems , respectively , have been discussed in detail in various reviews @xcite .
discussions on modifications owing to the possibility of andreev reflection appear in more recent studies @xcite , mostly focusing on the features of the quantum mechanical level density . in a similar fashion
, the aim of this paper is to determine how far a purely classical analysis may provide qualitative rationalization and quantitative predictions for the average quantum mechanical transport properties of a generic billiard such as that of fig .
[ fig : fig1 ] ; both in the presence or absence of andreev reflection . indeed , by performing exact calculations for the classical and quantum dynamics of low - energy quasiparticles
we find that the classical transport probabilities of electrons and holes , if appropriate , are in good quantitative agreement with the mean value ( to be defined below ) of the corresponding quantum mechanical scattering coefficients that determine the magnetoconductance of such systems .
while most of the previous works considered the case of zero or small magnetic field ( such that the classical dynamics is not altered ) , we particularly analyse the regime of finite magnetic field strengths and show that the classical trajectories which depend parametrically on the applied magnetic field suffice to describe the overall features of the observed non - monotonic behavior .
the article is structured as follows . in sec .
[ sec : s2 ] , after a brief discussion on the details of the studied system , we present precise numerical results of the magnetic - field dependence of the transport coefficients as determined by the quantum mechanical scattering matrix . in sec .
[ sec : s3 ] we first discuss the model describing the corresponding classical dynamics and provide an analysis for both the normal and the andreev version of the sinai billiard in sections [ sec : s3a ] and [ sec : s3b ] , respectively .
a synopsis is given in sec .
[ sec : s4 ] .
we consider ballistic transport of charge carriers in the @xmath2 sinai - billiard shown in fig .
[ fig : fig1 ] under an externally applied magnetic field .
the side length of the square cavity is taken @xmath3 where @xmath4 is the width of each of the leads attached to the left and right of the cavity .
the latter define source and sinks of ( quasi)particles . the central scattering disk possesses the radius @xmath5 , and it can be either a normal or a superconducting antidot . in the former case the antidot represents an infinitely high potential barrier while in the latter case it is considered as an extended homogeneous superconductor characterized by the property of andreev reflection @xcite .
experimentally , such antidot structures have been realized in periodic arrangements , thus forming superlattices @xcite .
the boundaries of the square cavity , numbered clockwise by the labels 1 through 4 , are always normal conducting potential walls of infinite height . in the presence of a superconductor
the quantum dynamics of the system can be described by the bogoliubov - degennes hamiltonian : @xmath6 where the diagonal operators determine the motion of particles and holes , respectively , and the off - diagonal elements take care of the coupling between particle- and hole - like excitations .
later , in our classical calculations we assume perfect andreev reflection meaning that _ all _ particles that hit the normal - superconducting ( ns ) interface are _ exactly _ retroreflected . in order to model this quantum mechanically
we have to consider perfect coupling between the normal - conducting region and the superconductor and simulate a bulk superconductor . to this end
, we take its size to be much larger than the superconducting coherence length @xmath7 ; @xmath8 is the fermi velocity . under these conditions , it is sufficient to consider a stepfunction - like behavior of the pair potential so that @xmath9 is constant inside the superconducting region and zero outside .
we also assume that the temperature is sufficiently smaller than the superconducting critical temperature so that @xmath10 does not diverge . for our numerical calculations we use a discretized version of the bogoliubov - degennes hamiltonian ( [ eqn : bdg_hamiltonian ] ) resembling the tight - binding approximation on a square lattice @xcite .
hence , @xmath11 becomes a matrix where only coupling between neighboring lattice sites is considered .
the submatrix @xmath12_{ij}$ ] has elements @xmath13 for @xmath14 and @xmath15 for nearest neighbors @xmath16 and @xmath17 .
the fermi energy @xmath18 is set to a value that allows six open channels in the leads .
the pairing matrix is given by @xmath19_{ij } = \delta_0\delta_{ij}$ ] if lattice point @xmath16 is inside the superconducting region and it is zero otherwise . to reproduce the correct dispersion relation in the continuum limit the onsite energies @xmath20 and the hopping energies @xmath15 have to fulfill the relation @xmath21 , where @xmath22 denotes a summation over nearest neighbors @xmath17 of site @xmath16 , see e.g. ref . .
in the presence of a magnetic field the hopping energies acquire a phase according to the peierls substitution @xcite , @xmath23 $ ] . here , @xmath24 is the vector potential , @xmath25 is the vector pointing from site @xmath17 to site @xmath16 and @xmath26 is the flux quantum . in general ,
the pair potential @xmath27}$ ] is also a complex number .
however , it can be chosen real ( @xmath28 ) if the vector potential @xmath24 is parallel to the screening currents near the ns interface @xcite .
this is achieved by choosing the symmetric gauge @xmath29 $ ] that accounts for a homogeneous magnetic field of strength @xmath0 in @xmath30 direction , perpendicular to the two - dimensional system . in what follows ,
we define as magnetic field unit the value @xmath31 for which the cyclotron radius is equal to @xmath4 .
the transport coefficients are calculated via a recursive decimation method , as explained in ref . .
this method enables the exact computation of the full scattering matrix @xmath32 , which yields scattering properties of quasiparticles with energy @xmath33 , incident on a phase - coherent structure described by a hamiltonian @xmath34 .
@xmath35 is the outgoing flux of quasiparticles along channel @xmath36 , arising from a unit incident flux along channel @xmath37 .
the quantum numbers @xmath36 indicating open scattering channels are conveniently written as @xmath38 , where @xmath16 indicates the leads , @xmath39 takes the discrete values @xmath40 and @xmath41 for particles and holes , respectively , and @xmath42 labels the quantum numbers associated with the quantization of the wavefunction in the transverse direction . as shown in ref .
, transport properties are determined by @xmath43 which is referred to as either a reflectance @xmath44 ( @xmath14 ) or a transmittance @xmath45 ( @xmath46 ) from quasi - particles of type @xmath47 in lead @xmath17 to quasi - particles of type @xmath39 in lead @xmath16 .
after normalization to unity with the number of open channels @xmath48 , @xmath49 , @xmath50 defines the andreev scattering probability , while for @xmath51 , it indicates a normal scattering probability .
such normalized quantities are equivalent to an angle - average and can be directly compared to the corresponding classical probabilities . in the remaining of the article we focus on the low - energy solutions of eq .
( [ eqn : bdg_hamiltonian ] ) with quasiparticle energy @xmath52 , which is appropriate for the model of perfect andreev reflection at the ns interface . in this case ,
particle and hole coefficients coincide .
hence , we adopt the shorthand notation @xmath53 , @xmath54 and @xmath55 , @xmath56 to indicate reflection and transmission probabilities , respectively .
due to interference effects , quantum scattering coefficients are rapidly oscillating functions of the fermi energy . therefore , in order to remove the quantum fluctuations , we perform an energy average over values @xmath57 $ ] , which corresponds to six open channels in the leads , @xmath58 $ ] , for each value of the magnetic field . the remaining parameters of our simulations are as follows . the width @xmath4 of the lead is @xmath59 . for the superconducting antidot , we define the pair potential via @xmath60 by choosing @xmath61 so that the diameter @xmath62 is approximately 6 times larger than the superconducting coherence length . to define the tunnel barrier in the case of the normal antidot
, an onsite potential of @xmath63 is added to all lattice sites lying inside .
note that the mesh lattice constant @xmath64 need not be defined explicitly if all energies are measured in @xmath65 yielding @xmath66 for every @xmath67-pair .
the above definitions are consistent with the requirements set in ref . about lengthscales , namely , @xmath68 and @xmath69 . here , @xmath70 is the fermi wavelength .
first we consider the sinai billiard with a normal antidot in the center acting as a potential barrier . in this case
coefficients with @xmath71 , i.e. , involving particle - to - hole conversion ( and vice - versa ) are identically zero as there is no andreev reflection at the antidot boundary .
particles ( holes ) can be either normally transmitted or reflected . in fig .
[ fig : trans_nc ] , the smoothed transmission is compared to the classical curve ( sec . [
sec : s3a ] ) revealing the same qualitative features .
even more remarkably , we see a very good quantitative agreement between both curves with deviations being within the amplitude of the small oscillations .
reflection is just symmetric to the transmission , i.e. @xmath72 , as both the classical and the quantum calculation respect unitarity .
( a ) ( b ) second we consider the case where the central antidot becomes superconducting .
andreev reflection now gives rise to non - zero @xmath73 and @xmath74 coefficients as shown in panel ( b ) of fig .
[ fig : trans_sc ] . upon comparison of the quantum results with the classical curves , we see again
that they agree very nicely .
the results are summarized in fig .
[ fig : trans_sc ] with vertical lines indicating two distinct values of the magnetic field , @xmath75 and @xmath76 , that are related to different qualitative features in the classical dynamics .
the largest differences occur for the particle - to - hole coefficients at the first critical field @xmath75 . there
, the classical transmission and reflection vanish abruptly , whereas the averaged quantum mechanical coefficients decay exponentially ( see inset of fig . [
fig : trans_sc ] ) .
however , we leave the analysis of such effects as well as the overall non - monotonic behavior with respect to the magnetic field for the next section for a discussion under the prism of the properties of the classical trajectories . to conclude this section
, we would like to show how the conductance , as an experimentally accessible quantity , changes when the antidot is made superconducting . in fig .
[ fig : cond ] , the magnetoconductance of a normal ( dots ) and for a superconducting ( solid line ) antidot is plotted . in the normal case
the linear - response low - temperature conductance is simply proportional to the transmission @xmath77 , according to landauer s formula @xmath78 .
et al _ @xcite have worked out generalizations for systems including superconducting islands or leads . for the andreev version of the sinai billiard system of fig .
[ fig : fig1 ] , the conductance is given by @xmath79 .
overall , we see that in the presence of andreev reflection the conductance of the system is larger than in the normal conducting case for magnetic fields @xmath80 . for larger fields
the particle - to - hole coefficients vanish and the conductances for both cases almost coincide .
in this section we study the classical dynamics of the incoming particles ( we focus on electrons but the same arguments apply to incoming holes ) for each of the two antidot structures described above .
the general form of the hamiltonian describing the dynamics of charged particles inside the cavity reads @xmath81 the index @xmath39 is used to describe the possibility that the propagating particles are either electrons ( @xmath40 ) or holes ( @xmath41 ) .
this generalization is necessary for a correct description of the dynamics in the setup with the superconducting antidot .
the canonical momentum vector is @xmath82 where @xmath83 is the mechanical velocity , the corresponding position vector being @xmath84 .
charge conservation yields @xmath85 for the effective masses and @xmath86 for the electric charge . the main property which distinguishes the two cases ,
i.e. , normal / superconducting antidot , is the interaction of the charged particle with the scattering disk .
the latter is captured by the elementary processes illustrated in fig .
[ fig : fig2 ] , namely , specular reflection ( sr ) versus the andreev reflection ( ar ) . in
what follows , we calculate the electronic transport properties by analyzing the ballistic propagation and escape of classical particles injected into the billiard via the opening pipe - like channels ( see fig .
[ fig : fig1 ] ) .
the initial conditions for incoming electrons are determined by the phase - space density @xmath87 \delta(m^*_e(v - v_f))\cos \theta , \\
\label{eq : density } \end{split}\ ] ] where @xmath88 $ ] is the angle of the initial electron momentum with the @xmath89-axis and @xmath90 and the coordinate origin is assumed at the center of the cavity .
the trajectories of the charged particles in the billiard consist of segments of circles with cyclotron radius @xmath91 ( with @xmath92 ) . at non - vanishing external field
the classical dynamics of both the normal and andreev billiards is characterized by a mixed phase space of co - existing regular and chaotic regions . at @xmath93
the superconducting antidot leads to an integrable dynamics , ( since trajectories are precisely retraced after retroreflection ) , while the corresponding normal device possesses a mixed phase space .
it is convenient to write the dynamics ( collisions with the walls and the antidot ) explicitly in the form of a discrete map . as the magnitude of the velocity remains constant in time ,
a simple parameterization of the dynamics is given by determining the position @xmath94 of the @xmath36-th collision with the boundary and the angle @xmath95 of the velocity vector with respect to the normal of the boundary at the collision point taken after the collision .
here , the term boundary refers to the walls 1 throughout 4 and the circumference of the antidot ( see fig . 1 ) .
there are three families of periodic orbits each forming a continuous set that occur in the classical dynamics and phase space of the _ closed system _
@xcite , i.e. without leads , leaving their fingerprints in the _ open system _ with the attached leads .
we will briefly discuss these periodic orbits in the following . at zero field
there are orbits bouncing between two opposite walls with velocities parallel to the normal of the corresponding walls . at finite but weak @xmath0-field strength
the periodic orbits form a rosette and incorporate collisions with the antidot and the walls .
these periodic orbits are typical , i.e. dominant up to a critical field value @xmath76 . for magnetic fields above
@xmath76 the cyclotron radius is so small that no collisions with the antidot can occur and skipping orbits , describing the hopping of the electrons along the billiard walls , become dominant .
all periodic orbits possess an eigenvalue one of their stability matrix @xcite and all periodic orbits possess unstable directions .
we remark that the above - discussed periodic orbits of the closed billiard are not trajectories emerging from and ending in the leads of the open billiard .
however , trajectories of particles coupled to the leads ( i.e. , injected and transmitted / reflected ) can come close to the periodic orbits of the open billiard thereby tracing their properties . this way
the presence of the periodic orbits reflects itself in the transport properties .
first we consider the transport of electrons through the sinai billiard ( fig .
[ fig : fig1 ] ) with a normal antidot .
the relevant quantities determining the current flow through the device are the transmission @xmath77 and reflection @xmath96 coefficients for electrons defined as the percentage of the initial electrons leaving the device from the right and left lead , respectively .
additional quantities that are helpful for an understanding and analysis of the system dynamics are the mean number of collisions per incoming electron with the walls ( 1 - 4 ) , @xmath97 , and with the antidot , @xmath98 .
we calculate these quantities by numerical simulation for different values of the external magnetic field @xmath0 .
it is convenient to use a dimensionless form of the classical equations of motion by employing the scaling @xmath99 and @xmath100 for the spatial coordinates and @xmath101 ( with @xmath102 ) for the time coordinate .
the above quantities are calculated for 100 values of the magnetic field strength varying from 0.02 to 2 using an ensemble of @xmath103 different initial conditions distributed according to eq .
( [ eq : density ] ) for each @xmath0-field value . the magnetic field dependence of the coefficients @xmath77 and @xmath96
is shown in fig .
[ fig : fig3]a .
the obtained curves are quite irregular , possibly indicating the presence of fractal fluctuations in the magnetoconductance of the system @xcite .
( a ) classical electron transmission , @xmath77 , and reflection , @xmath96 , coefficients for the normal antidot device of fig .
[ fig : fig1 ] as a function of the applied magnetic field @xmath0 .
( b ) the mean number of collisions with the boundary of the square cavity ( walls 1 - 4 ) , @xmath97 , and with the circumference of the antidot , @xmath98 , as a function of @xmath0 .
the field strength is given in units of @xmath104 . ] in fig .
[ fig : fig3]b , we present the parametric dependence of the mean quantities @xmath97 and @xmath98 .
interestingly , the mean number of collisions with the walls remains constant @xmath105 for almost all values of the field strength .
this means naturally that also the accumulated number of collisions of all injected trajectories with the walls is independent of the field strength for the whole regime considered .
this number is obtained by integrating the occupancy of the trajectories in phase space , i.e. , their measure , over all possible velocities and the boundary of the cavity ( defined by the walls 1 - 4 including the leads ) .
its invariance with respect to the field strength is of combined geometrical and dynamical origin and can be understood as follows .
the escape probability is well approximated by @xmath106 where @xmath107 is the measure of phase space points on the left and right leads visited by the escaping electrons , while @xmath108 is the total measure involved in the dynamics of the system along the boundary defined by the walls 1 - 4 ( including the leads ) .
the corresponding integrals can be estimated as @xmath109 and @xmath110 .
here , @xmath111 is the mean phase space density on the leads and @xmath112 is the mean phase space density on the entire boundary , both integrated over the momenta . due to the symmetric setup of the leads relative to both the @xmath89 and @xmath113 axis
, we have @xmath114 .
thus , @xmath115 and the mean number of collisions with the wall is @xmath116 .
the behavior of @xmath98 is more complicated because the dynamical occupation of the antidot s circumference strongly depends on the value of the external field .
one can clearly distinguish three regimes : ( i ) the low field region ranging from @xmath117 to @xmath118 , ( ii ) the intermediate field region with @xmath119 , and ( iii ) the high field region with @xmath120 .
all three regions are characterized by different properties of the corresponding phase space .
these are revealed by the study of the phase space structure using poincar surfaces of section ( psos ) for different values of the applied field @xmath0 .
we employ @xmath121 sections defined by the condition @xmath122 .
it turns out that all calculated surfaces of section reveal a mixed phase space . to further quantify our analysis we calculate the relative weight @xmath123 of those trajectories on the psos that exhibit collisions with the antidot . since collisions with the antidot are the only possibility to obtain dynamics that is sensitive with respect to the initial conditions , @xmath123 is also a measure for chaoticity in phase space @xcite .
we partition the energetically allowed phase space on the @xmath121 plane into @xmath124 cells of equal size and define on each cell @xmath125 the characteristic function @xmath126 as : @xmath127 we subsequently approximate @xmath128 .
the function @xmath129 is shown in fig .
[ fig : fig4 ] . the initial plateau at @xmath130 shows clearly that the system is to a large portion chaotic for low magnetic fields .
this explains the fact that in this range of fields the mean number of collisions with the antidot @xmath98 is almost constant .
the degree of chaos in the phase space of the system is large enough thereby ensuring that , with the exception of trajectories of negligible measure , each trajectory hits the circumference of the antidot .
hence , as evaluated in a similar fashion to @xmath108 , the total measure of phase space points @xmath131 involving the circumference is equal to @xmath132 . following the arguments given above for @xmath133
, we estimate @xmath98 as @xmath134 , yielding @xmath135 , which is in very good agreement with fig .
[ fig : fig3]b . in the intermediate field region
the weight of the chaotic trajectories decreases almost linearly and becomes vanishingly small in the high field region .
the linear decrease of @xmath123 leads to a linear decrease of @xmath98 for this range of magnetic fields .
above @xmath76 the cyclotron radius of the electron trajectories is so small ( @xmath136 ) that no collision with the antidot is possible , i.e. , @xmath137 .
relative weight of the chaotic part @xmath123 of the phase space as a function of the applied field @xmath0 ( in units of @xmath104 ) for the normal conducting sinai - shaped billiard of fig .
[ fig : fig1 ] . ]
we can now understand the non - monotonic behavior of the functions @xmath138 , @xmath139 by considering the representative trajectory dynamics for various @xmath0-values . according to fig .
[ fig : fig3]a the low field region possesses two subregions : ( i ) @xmath140 and ( ii ) @xmath141 .
similarly the intermediate field region can be divided to : ( i ) @xmath142 and ( ii ) @xmath143 .
note that the points defining the magnetic field windows , @xmath144 and @xmath145 , also mark qualitative changes in the functions @xmath146 and @xmath129 . at very low fields , @xmath147 , we observe that @xmath96 is slightly larger than @xmath77 .
this owes to many trajectories exhibiting only one collision with the antidot and reflected directly back to the lead from where they came .
the typical configuration consists of an incoming electron moving almost as a free particle , hitting the antidot once , and escaping from the billiard to the left lead ( electron reflection ) .
otherwise , in the low field region ( i ) the main process is the transmission of electrons . as the magnetic field increases , the reflection angle at the circumference of the antidot increases too and the incoming electron , after hitting the antidot , suffers two or more collisions with some of the walls 1,2,3 or 4 before escaping to the right lead ( electron transmission ) .
this mechanism , along with a significant amount of electrons that are initially emitted with a larger angle ( @xmath148 ) , hitting directly the upper or lower wall of the cavity ( or even its right wall ) suffering specular reflection and exiting to the right lead of the device , establishes electron - transmission as the main process in the low field region .
however , it is evident from fig .
[ fig : fig3]a that the difference between @xmath77 and @xmath96 is quite small
. there are complex trajectories with more than 10 collisions with the walls and 5 - 8 collisions with the antidot possessing a finite measure in phase space .
these give a non - vanishing contribution to @xmath96 thereby maintaining its mean value around @xmath149 .
in fact @xmath77 and @xmath96 fluctuate insignificantly around their mean values ( @xmath150 for @xmath77 and @xmath149 for @xmath96 ) . as the fields increases above @xmath151
, trajectories with a larger number of collisions with the walls may become statistically more important but as long as @xmath152 the trajectories with several collisions with the antidot have still a significant measure ( see figs . [
fig : fig4 ] and [ fig : fig3]b , respectively ) .
overall , for @xmath141 , the main process is the reflection of electrons yielding a large difference between @xmath77 and @xmath96 .
typical trajectories have one or two collisions with the walls and a single collision with the antidot .
the incoming electron hits the antidot once , is specularly reflected and escapes from the billiard to the left lead , after suffering one more collision with the wall 4 . at intermediate fields
, we observe an almost monotonic decrease of @xmath96 and an increase of @xmath77 owing to the combined decrease of @xmath146 and @xmath129 . the small plateau feature in fig .
[ fig : fig4 ] around @xmath153 is also reflected in the change of slope in the transport probabilities within subregion ( i ) . at its upper limit , @xmath154 , the two coefficients become equal , @xmath155 .
most trajectories have @xmath156 collisions with the walls and @xmath157 collisions with the antidot . in window
( ii ) of the intermediate - field region with @xmath158 , the process of transmission is much stronger than the process of reflection .
most trajectories have a few ( @xmath159 ) collisions with the walls and no collision with the antidot .
a typical trajectory of the incoming electron , due to the small cyclotron radius , suffers one collision with wall 1 , three collisions with wall 4 , one collision with wall 3 ( 5 collisions with the walls in total ) and then escapes from the cavity through the right lead ( electron - transmission ) .
there is also the case in which the incoming electron misses the right lead of the device , and after suffering many collisions with the walls finally escapes to the left lead , contributing to the process of reflection .
the same scenario is valid also for the high field region .
compared to the dynamics of the cavity with the normal antidot , the andreev sinai billiard with the central superconducting disc exhibits basic differences due to the occurrence of trajectories which suffer andreev reflection , instead of specular reflection , at the circumference of the antidot .
first , the complete description of the transport properties of the system requires the introduction of two additional coefficients describing electrons that escape as holes either to the left or the right lead ; @xmath73 ( reflection ) and @xmath74 ( transmission ) , respectively . in figs .
[ fig : fig5]a and [ fig : fig5]b , all probabilities are plotted as a function of the applied field @xmath0 . in fig .
[ fig : fig5]c , we present the quantities @xmath160 and @xmath146 following the definition of sec . [
sec : s3a ] .
@xmath161 and @xmath162 ( with @xmath163 ) exhibit irregular fluctuations as a function of @xmath0 , similar to those obtained for the normal billiard .
the function @xmath97 is almost identical to the corresponding function obtained for the normal case .
qualitatively , the mean number @xmath98 of collisions with the superconducting antidot is also similar to that in fig .
[ fig : fig3]b . however , while @xmath76 is remaining the same , the value of the critical field @xmath75 is shifted to the larger value @xmath164 .
the former should be expected since it does not involve any collisions with the superconducting disc .
the difference in @xmath75 is explained by calculating the relative weight of the chaotic trajectories as in sec .
[ sec : s3a ] .
the result is shown in fig .
[ fig : fig6 ] . for magnetic fields up to @xmath165 a large part of the phase space of the system is chaotic ensuring the equal mean phase space density on the boundary of the square cavity and the circumference of the antidot . for @xmath166 an almost linear decrease of
@xmath123 leads to a corresponding linear decrease of @xmath98 . at @xmath167 the chaotic part of the phase space vanishes due to the fact that no collisions with the defocusing perimeter of the antidot are possible .
an additional peculiarity of the superconducting device appears for the intermediate field region , @xmath168 : _ all _ possible trajectories possess an even number of collisions with the antidot , yielding a vanishing transmission and reflection of holes , @xmath169 . this interesting feature is related to the generic properties of andreev reflection and is analyzed in appendix a. classical results for the billiard of fig . [
fig : fig1 ] with a superconducting disc in the center , as a function of the applied magnetic field @xmath0 ( in units of @xmath104 ) .
( a ) classical electron transmission , @xmath77 , and reflection ( @xmath96 ) coefficients .
( b ) classical electron - to - hole ( andreev ) transmission , @xmath74 , and reflection , @xmath73 , probabilities .
( c ) mean number of collisions , @xmath97 , with the boundary of the square cavity ( walls 1 - 4 ) and with the circumference of the antidot , @xmath98 . ]
let us now consider the transport coefficients in more detail .
following the qualitative features in figs .
[ fig : fig5]a and [ fig : fig5]b , we divide the low field region into four windows : ( i ) @xmath170 , ( ii ) @xmath171 , ( iii ) @xmath172 and ( iv ) @xmath173 .
note that different regimes roughly coincide with different qualitative features of the functions @xmath146 and @xmath129 , as indicated previously .
relative weight of the chaotic part of the phase space @xmath123 as a function of the applied field @xmath0 ( in units of @xmath104 ) for the andreev version of the billiard in fig .
[ fig : fig1 ] . ] in region ( i ) , due to the large cyclotron radius ( almost vanishing curvature ) most trajectories have a single collision with the antidot .
in fact , at very small fields , @xmath174 , there is practically no collision with the walls .
the typical process consists of an incoming electron moving almost on a straight line hitting the antidot once , being converted into a hole which nearly retraces the electron path moving towards the left lead and finally leaving the device .
therefore , @xmath73 in this region is larger than @xmath74 , @xmath77 and @xmath96 .
there is however a significant amount of electrons emitted initially with a large enough angle , @xmath148 , which hit directly the upper or lower wall of the device suffering normal reflection and then exiting to the right lead .
this process gives a finite electron transmission coefficient , yielding a significant contribution to @xmath77 .
for even larger emission angles , there is a second but less significant set of trajectories that exhibit specular reflections at the walls 2 , 3 , and 4 .
particles along these paths escape finally from the cavity through the left opening , overall leading to very small values of @xmath96 . with increasing @xmath0 ,
the curvature of the trajectories also increases so that an additional collision with the wall takes place .
the incoming electron hits again the antidot , becoming a hole and as the curvature is increased the hole can not escape from the narrow left lead , subsequently hitting wall 1 , being specularly reflected and leaving the device from the right lead . therefore , @xmath74 is increased at the cost of @xmath73 . in region ( ii )
, the curvature of the trajectories increases further .
a typical trajectory for an incoming electron , after being andreev reflected at the antidot , hits the wall 4 and escapes to the left lead after specular reflection . as a result
@xmath73 increases in this region .
we also observe a decrease of @xmath77 owing to the increased curvature of the trajectories making it difficult for the incoming electrons to avoid the collision with the antidot and it is therefore harder to encounter the outgoing right lead . in region
( iii ) , there is no significant variation of the observables @xmath175 and @xmath176 . as the intermediate field regime defined by the linear decrease in @xmath98
is approached , in ( iv ) we observe a sudden decrease of both @xmath74 and @xmath73 with a simultaneous increase of , predominantly , @xmath77 .
although the measure @xmath123 of trajectories with several collisions with the antidot is non - zero , thereby , contributing to the chaotic part of the phase space , the predominant part of the trajectories do not experience andreev reflection . in the regime with @xmath177 , we encounter @xmath169 due to the fact the every trajectory has an even number of collisions with the antidot ( see appendix a ) .
the predominant part of the orbits have 5 to 7 collisions with the walls 1,3 and 4 thereby hopping along the boundary of the square cavity .
finally , the high - field region is characterized by @xmath169 , and @xmath137 as in the case of the normal antidot .
performing simulations of the classical and quantum dynamics of low - energy quasiparticles , we showed that a purely classical analysis may be used as an interpretation tool for the average transport properties of generic normal and andreev billiards .
in particular , the parametric dependence on the strength of a perpendicular magnetic field b was studied . as the strength increases
, this dependence of the classical trajectories on the applied magnetic field drives the classical dynamics from mixed to regular for both types of billiards .
the latter is grossly reflected in the non - monotonic behavior of the magnetoconductance at intermediate fields .
owing to the increasing trajectory bending , a slight increase of the conductance @xmath178 at small fields of the normal billiard is followed by a significant valley whose minimum defines the passage to intermediate strengths .
scattering of particles with the sinai - billiard disc starts reducing , also triggering the relative weight @xmath123 of the chaotic part of phase space to shrink .
around @xmath104 , which corresponds to a cyclotron radius equal the disc radius , skipping orbits settle in and transport properties converge towards the high - b limit . turning on the superconductivity at the sinai - billiard disc results in the interplay of the bending of the trajectories and the occurring particle - to - hole conversion .
the magnetic field drives the integrable correlated motion of particles and holes into a mixed dynamics regime evident by an initial tendency of @xmath179 towards @xmath180 at small fields , typical for systems with phase space having a relative large chaotic part .
compared to the normal case , increasing @xmath0 andreev reflection counteracts to the reduction of @xmath123 , which occurs eventually but at higher field strengths .
hence , realizing the andreev billiard leads to qualitatively different behavior . rather than a magnetoconductance dip
, we note a reentrance effect of the conductance towards its initial higher value .
our classical calculations provide not only a qualitative rationalization of the observed properties of the exact quantum mechanical scattering coefficients and of the corresponding magnetoconductance spectrum but also allow us to make quantitative predictions , as evidenced by the remarkable agreement between the classical and quantum values .
this is ideal for the designing of experimental setups and a simple analysis of the results , since classical simulations are much less time consuming . in the present paper we studied the ( energy )
averaged transport properties , i.e. , removing the quantum fluctuations .
yet the averaged quantum results contain weak localization effects at zero and small magnetic fields .
these quantum corrections to the averaged transmission are of order one ( more precisely -1/4 for chaotic ballistic systems ) compared to the classical contribution which is proportional to @xmath48 .
more specifically , for the transmission per channel for the normal conducting sinai billiard with @xmath181 the negative quantum correction at zero field is expected to be below 0.1 , in line with the numerical results depicted in fig .
[ fig : trans_nc ] .
it is indeed possible to extend the existing semiclassical theory for ballistic weak localization @xcite to averaged quantum transport through andreev billiards @xcite .
however , our focus was on the finite @xmath0-field range where weak localization effects do not exist .
rather , conductance fluctuations in this regime encode additional quantum information , and previous results in closed and open billiards ( either with a normal or a superconducting lead ) indicate that such fluctuations are interwined with the underlying classical properties .
therefore , we envisage that our study could be further developed and utilized both theoretically and experimentally in future investigations .
gf acknowledges funding support by the science foundation ireland .
al and ms were supported by the _ deutsche forschungsgemeinschaft _ within the research school grk 638 .
[ t ] numerical results show that for magnetic fields @xmath182 there are only trajectories with an even number of collisions with the superconducting antidot , yielding vanishing transport coefficients @xmath74 and @xmath73 . in order to derive the first critical field it is helpful to consider the mapping of the guiding centers of the trajectory arcs in the presence of andreev reflection .
[ fig : aref ] shows two segments of an orbit right before and right after a collision with the ns - interface .
the center of the second arc can be constructed from the center of the first one via point reflection at the collision point at the ns - interface .
consider one segment of an orbit with radius @xmath183 reaching from the outer wall to the antidot .
the distance between the center of this arc and the outer wall has to be less than @xmath183 and the same holds for the distance between the center and the antidot .
this means that the center has to be inside the shaded area shown in fig .
[ fig : sinai ] . consider an orbit that hits the superconductor only once .
such an orbit has to reach from the outer wall to the antidot and back to the wall . so the centers before and after the andreev reflection have to be located inside the shaded area .
this condition is easiest to fulfill for an orbit that has its center at one midpoint of the inner quadratic boundary , as shown in fig .
[ fig : sinai ] .
the question is if for a certain radius the center after the andreev reflection can no longer be mapped into the shaded region .
the distance @xmath184 between the antidot and the inner square is @xmath185 , where l is the side length of the square cavity and @xmath186 is the radius of the antidot . applying the cosine - theorem to the triangle shown in fig .
[ fig : sinai ] we get @xmath187 .
so the angle @xmath188 can be written as @xmath189 the distance between the final point and the central horizontal line is @xmath190 . in order to get only one single andreev reflection ,
the final point has to be inside the shaded region , therefore we have to claim @xmath191 , which means @xmath192 solving this inequality for @xmath193 and @xmath194 we find a critical radius @xmath195 , which corresponds via @xmath196 to a critical field of @xmath197 up to now we have only shown that for a magnetic field @xmath182 the particles hit the superconductor at least twice consecutively .
but now it is easy to see that the number of andreev reflections is indeed even .
two andreev reflections in a row correspond to a rotation of the center around the center of the antidot , as shown in fig .
[ fig : rot ] .
after an even number of andreev reflections the center of the arc is always located on the outer dashed circle , which has a radius greater than @xmath198 .
after an odd number of reflections at the superconductor it is on the inner dashed circle with a radius smaller than @xmath198 . therefore the particle can only ` escape ' the superconductor after an even number of collisions , which explains the fact .
for a chaotic andreev billiard one finds for the weak localization correction to the averaged conductance @xmath199 , where @xmath200 is the size of the interface with the superconductor and @xmath4 is the width of the leads attached , see : a. lassl , diploma thesis , universitt regensburg , unpublished , 2003 . | we perform a comparative study of the quantum and classical transport probabilities of low - energy quasiparticles ballistically traversing normal and andreev two - dimensional open cavities with a sinai - billiard shape .
we focus on the dependence of the transport on the strength of an applied magnetic field @xmath0 . with increasing field strength the classical dynamics changes from mixed to regular phase space .
averaging out the quantum fluctuations , we find an excellent agreement between the quantum and classical transport coefficients in the complete range of field strengths .
this allows an overall description of the non - monotonic behavior of the average magnetoconductance in terms of the corresponding classical trajectories , thus , establishing a basic tool useful in the design and analysis of experiments . |
intermediate - mass stars , with initial masses @xmath1 , are thought to contribute significantly to the synthesis of c , n , o , and ne through the cno cycle and he burning .
these nuclear fusion products are ejected through mass loss as the stars evolve from their agb ( asymptotic giant branch ) phase into planetary nebulae ( pne ) .
hence , a pn can be regarded as a messenger bearing information on the nucleosynthesis within these stars . however , the optically - visible material in pne represents matter accumulated over the pn lifetime , making it difficult to extract information on , for instance , pure he - burning products just by observing the pne shells .
soft x - rays , detected from several pne , are thought to originate in hot plasmas produced by shocks in fast stellar winds @xcite .
these fast winds develop during the later evolutionary stages of the central star , so the x - rays are thought to be emitted by the star s late - phase products .
x - ray spectroscopy of pne will thus allow us to diagnose products of a particular nucleosynthesis phase inside intermediate - mass stars .
however , it has been difficult to resolve k - lines from c , n , and o , the major cno and he - burning products , because the line energies were too low for x - ray ccds .
in addition , these objects are often too x - ray faint for high - resolution spectroscopy using gratings , even when their angular size is small enough .
the 5th japanese x - ray satellite _ suzaku _ ( _ astro - e2 _ ; mitsuda et al .
2004 ) was developed by a japan - us collaboration as a successor to _ asca _ , and was launched successfully on 2005 july 10 .
its x - ray imaging spectrometer ( xis ) comprises four ccd cameras , xis-0 through xis-3 @xcite .
three of the cameras use front - illuminated ( fi ) ccd chips while xis-1 utilizes the back - illuminated ( bi ) technology .
compared to similar instruments on preceding x - ray missions , both the bi and fi chips have better responses to x - rays with energies below @xmath2 kev : their fwhm energy resolution at 0.5 kev is @xmath3 ev ( fi ) and @xmath4 ev ( bi ) , with insignificant low - energy tails .
we observed the pn , bd + 30@xmath0 3639 , with _
suzaku _ in order to resolve the anticipated carbon and oxygen emission lines .
since its detection using _ rosat _
@xcite , this object has been known as the x - ray brightest pn , emitting spatially - extended soft x - rays from inside its @xmath5 diameter optical shell @xcite . the x - ray emitting plasma
is thought to have highly non - solar abundance ratios , as shown by the strong ne - k line first detected using _
@xcite , and a broad sub - kev spectral hump , presumably a blend of c , n , and o lines , subsequently detected with _
chandra _ @xcite .
the present observation of bd + 30@xmath0 3639 was performed for a net exposure of 34.4 ksec on 2005 september 20 , ( ) , as part of the initial performance verification of _
suzaku_. both the xis and the hard x - ray detector ( hxd ; @xcite ) onboard were operated in their nominal modes .
we do not however use the hxd data , since its non - imaging @xmath6 field of view also contains part of the large supernova remnant , g65.2 + 5.7 , which overlaps with the target pn .
the target was clearly detected in all xis cameras ; figure [ szimage ] shows 0.30.7 kev images obtained with two of them .
we define the source region as a @xmath7 radius circle centered on the source , which encloses @xmath8 of signal photons from the pn that is essentially a point source at the angular resolution of the _ suzaku _ x - ray telescopes .
the background region is defined as a surrounding annulus with an outer radius of @xmath9 , as indicated in figure [ szimage]a .
after background subtraction , net signal counts are @xmath10 , @xmath11 , @xmath12 , and @xmath13 from xis-0 , 1 , 2 , and 3 , respectively .
since the three fi cameras ( xis-0 , 2 , 3 ) are essentially identical , we co - add their data , and refer to them as `` xis-023 '' below .
as shown in figure [ x1spec ] , the on - source xis-1 spectrum clearly reveals at 0.37 kev the k@xmath14 line from hydrogenic carbon ( c vi ) .
its absence in the background spectrum ensures that the line is coming from the pn itself .
the spectra also show k@xmath14 lines from o vii , o viii , and ne ix .
some of these lines are found in the background spectrum as well , but with much reduced intensities ; these background lines are thought to come from diffuse galactic soft x - ray components which may well extend to the present galactic latitude of @xmath15 , possibly with additional contribution from g65.2 + 5.7 .
at a temperature of @xmath16 kev @xcite , the c vi to o vii ( the triplet summed ) line emissivity ratio of a solar - ratio plasma is @xmath17 . as illustrated in figure [ x1spec ] in green ,
this ratio will be reduced to @xmath18 @xcite due to interstellar plus intra - nebular absorption ( with the latter probably dominant ; @xcite ) , and the decrease in xis-1 efficiency toward lower energies .
nevertheless , the xis-1 spectrum yields comparable numbers of c vi and o vii line photons , suggesting a highly enhanced c / o ratio of the x - ray emitting plasma .
this early in the mission , the xis calibration is still being fine - tuned .
there are uncertainties in the energy gain , and also in the low - energy efficiency which has decreased since the start of observations .
we solved the former by self - calibrating ( within @xmath19 ) the gains using the emission lines from bd + 30@xmath0 3639 themselves .
the latter effect is attributed to excess absorption due to the build - up of contaminant in front of the xis cameras .
we estimated the chemical composition of the excess absorber using a _ suzaku _ observation of the isolated neutron star rx j1856.6 - 3574 @xcite , which is known to exhibit a blackbody with a temperature of 63 ev , absorbed by a low column of @xmath20 @xmath21 @xcite . to accurately determine the contaminant thickness for xis-1 and ( separately ) xis-023 , we included in all analysis a simultaneous fit to the archival ( ) background - subtracted _ chandra _ acis - s spectrum of the same object .
we jointly fit the background - subtracted xis-1 , xis-023 , and _
acis - s spectra , using an absorbed single - temperature vapec model @xcite .
the abundances ( relative to solar ; @xcite ) of c , n , o , ne , and fe , were left free ; that of he fixed to solar ; other heavy elements were neglected ; @xmath22 and the intervening ( interstellar plus intra - nebular ) hydrogen column density @xmath23 were left free . as shown in figure [ combined - fit ] , the model gives a reasonable joint fit to the three spectra , with @xmath24 .
we determined @xmath25 kev and @xmath26 .
figure [ combined - fit ] also illustrates the difference in the energy - resolving power between the xis and acis - s .
this fitting procedure can not in reality constrain the absolute ( i.e. , relative to hydrogen ) metal abundances , since the low - energy flux from a plasma with @xmath27 kev is dominated by emission lines from the metals themselves , even when the metallicity is @xmath2 solar .
furthermore , the continua , due to hydrogen , ( presumably non - solar abundance ) helium , and the enhanced metals themselves , can not be independently determined .
nevertheless , we expect abundance _ ratios _ among metals to be reasonably constrained by the lines which are individually resolved with the xis . to see this , we produced in figure [ contour ] confidence contours for the c and o abundances using the joint vapec fits . thus , the contours are highly elongated along the proportionality between the two quantities , and the c / o ratio is successfully constrained as @xmath28 c / o @xmath29 solar at 90% confidence with the best estimate at 85 .
( the range does not change significantly even allowing @xmath23 to float . ) in the same way , we have obtained @xmath30n / o @xmath31 and @xmath32 ne / o @xmath33 , with the best fits of 3.2 and 5.8 , respectively , and fe / o@xmath34 , all in solar units .
below , these results are examined for various systematic effects .
we have so far assumed the he / h ratio to be solar .
however , increasing it to 5 , 10 , or 20 solar does not affect the relative c / n / o / ne abundances by more than 5% .
similarly , the results remain unchanged within @xmath35% when the thickness of the xis excess absorber is varied by @xmath36% around the best estimates ( equivalent to @xmath37 @xmath21 ) , or the background spectra are derived from several other regions ( e.g. , using larger annuli or using limited azimuthal sectors ) in the xis field of view . to relax the assumption of isothermality , we refitted the three spectra jointly with a sum of two vapec components , which are constrained to share common abundances and the same @xmath23 .
we tentatively fixed the c / o ratio at various trial values , and let the two temperatures vary freely .
even with this extra degree of freedom , acceptable fits were reproduced only when the assumed c / o ratio is in the range 35120 solar .
the two temperatures were typically found at @xmath38 kev and @xmath39 kev .
further , the c / o ratio stayed above @xmath40 even when fitted with a model allowing non - equilibrium ionization conditions .
the largest systematic effect on our results comes from uncertainty in @xmath23 , because our best - fit value ( @xmath41 @xmath21 ) is twice that estimated from extinction data at longer wavelengths @xcite .
we therefore repeated the joint , single - vapec fit to the three spectra , but this time fixing @xmath23 at @xmath42 @xmath21 .
then , @xmath22 increased to @xmath43 kev , and the c / o ratio decreased by a factor of @xmath44 , while the ne / o ratio remained nearly unchanged .
these new parameters are consistent with the recent _ chandra _ letg results on bd + 30@xmath0 3639 @xcite .
however , the ( non - reduced ) @xmath45 for the joint fit to the _ suzaku _ xis and _ chandra _ acis data increases by @xmath46 when @xmath23 is forced to take the smaller value .
this suggests that the x - ray absorption is actually higher than is implied by the extinction data , perhaps because the intra - nebular absorber is metal - enriched .
based on these evaluations , we take as our result the original single - vapec solution ( i.e. , c / o @xmath47 ) with free @xmath23 , with the reservation that the c / o ratio could be subject to a systematic over - estimation by a factor of @xmath44 .
this factor reflects possible non - isothermality , non - equilibrium ionization conditions , or the uncertainty in @xmath23 . in any event
, the very high carbon over - abundance relative to oxygen is a robust result , confirming the previous suggestions @xcite .
the three spectra analyzed consistently indicate an absorbed 0.22.0 kev flux of @xmath48 erg s@xmath49 , in agreement with previous measurements @xcite .
correction for the absorption by @xmath50 @xmath21 increases this flux by a factor of @xmath51 , yielding an absorption - corrected 0.22.0 kev luminosity of @xmath52 ergs s@xmath49 at a distance of 1.3 kpc @xcite .
( if the fit with @xmath53 @xmath21 is used , the absorption - corrected quantities become smaller than these by a factor of @xmath44 . )
this is only @xmath54 of the kinetic luminosity , @xmath55 ergs s@xmath49 , supplied by the fast stellar wind @xcite .
this , together with the x - ray emitting region just fitting inside the optical shell @xcite , supports the interacting wind model @xcite as a scenario to explain the x - ray emission from this pn .
observations in the optical and neighboring wavelengths give the nebular abundance ratios of bd + 30@xmath0 3639 as c / o@xmath56 , n / o@xmath57 , and ne / o @xmath58 in solar units @xcite .
compared with these , our x - ray results imply qualitatively similar , but much more extreme , abundance patterns .
this is not surprising , since the x - ray emitting plasma is considered to represent a very limited radial zone of the stellar interior .
in fact , assuming , based on the _ chandra _ image , that the x - ray emitting plasma fills a sphere of radius 0.01 pc ( @xmath59 ) , its mass is estimated as @xmath60 from the observed emission measure of @xmath61 @xmath62 .
this can be supplied only in @xmath63 yr by the mass loss of @xmath64 yr@xmath49 from the central star @xcite .
which part of the stellar interior provides the x - ray emitting plasma ?
the most outstanding feature is the extreme carbon enhancement over oxygen , by a factor of @xmath3 in the _ number _ ratio , or @xmath40 in the _ mass _ ratio .
this is a typical value expected from competition between the triple-@xmath65 and @xmath66c(@xmath67)@xmath68o reactions during he shell - burning flashes @xcite , for an initial mass of @xmath69 .
therefore , we presume that the x - rays come primarily from products of he shell - burning episodes which took place in the agb phase
. the possible factor - of - three uncertainty will not affect this inference significantly , since the theoretical calculations are thought to be uncertain to a similar degree .
the observed high ne / o ratio , in contrast , requires a neon - producing path without @xmath68o .
presumably @xmath70n , concentrated through the cno cycle , burnt as @xmath70n(@xmath67)@xmath71f(@xmath72)@xmath71o(@xmath67)@xmath73ne during the he shell flashes .
this produces @xmath73ne , which is indistinguishable in x - rays from the more abundant @xmath74ne .
finally , the hint of enhanced n / o ratio can be explained if some fraction of the nitrogen from the cno cycle , stored in the upper he layer , escaped without being caught up in the he shell burning .
these considerations lead us to the following scenario for this pn .
the hydrogen - rich envelope is likely to have already been expelled .
the outer parts of the he layer , composed of radiative and convective zones , are currently being ejected in the form of fast stellar winds with a velocity of @xmath75 km s@xmath49 .
the winds carry the abundant carbon and @xmath73ne produced in he shell flashes @xcite , as well as unprocessed nitrogen .
these ejecta are shock - heated to emit x - rays .
a more quantitative comparison with the optical spectra of the central stars of pne @xcite will provide a valuable calibration of the models of stellar evolution and nucleosynthesis .
such work may also allow us to estimate the initial progenitor mass , and to clarify when the fast stellar winds start blowing . in closing , the present _ suzaku
_ observation has provided a valuable opportunity of witnessing the carbon nucleosynthesis inside an evolved star , which is one of the central building blocks of the modern astronomy but has rarely been observed directly .
mm , together with the co - authors , expresses her deepest gratitude to the _ astro - e / astro - e2/suzaku _ project members .
authors thank prof .
masayuki fujimoto and dr .
takuma suda of hokkaido university for elucidating discussion .
they are also grateful to dr .
joel kastner of rochester institute of technology for communicating their latest _ chandra _ letg results on bd + 30@xmath0 3639 . | bd + 30@xmath0 3639 , the brightest planetary nebula at x - ray energies , was observed with _ suzaku _ , an x - ray observatory launched on 2005 july 10 . using the x - ray imaging spectrometer , the k - lines from c vi , o vii , and o viii were resolved for the first time , and c / o , n / o , and ne / o abundance ratios determined .
the c / o and ne / o abundance ratios exceed the solar value by a factor of at least 30 and 5 , respectively .
these results indicate that the x - rays are emitted mainly by helium shell - burning products . |
recent advances in extragalactic astrophysics show the close link of a disk galaxy dynamical evolution and its chemical and photometric behaviour over the hubble timescale . in spite of remarkable succes of the modern theory of galaxy chemical evolution in explaining the properties of evolving galaxies ( @xcite ) its serious shortcomings concern the multiparameter character and practical neglecting of dynamical effects .
the inclusion of simplified dynamic into the chemical network ( @xcite ) and vice versa the inclusion of simplified chemical scheme into the sophisticated 3d hydrodynamical code ( @xcite ) gives very promising results and allows to avoid a formal approach typical to standard theory . in this paper
the interplay between a disk galaxy dynamical evolution and its chemical behaviour is studied in a frame of a simplified model which provides a realistic description of the process of galaxy formation and evolution over the cosmological timescale .
the evolving galaxy is treated as a system of baryonic fragments embedded into the extended halo composed of dark nonbaryonic and baryonic matter .
the halo is modelled as a static structure with dark ( dh ) and diluted baryonic ( bh ) halo components having plummer type density profiles ( @xcite ) : @xmath0 and @xmath1 where @xmath2 @xmath3 the dense baryonic matter ( future galaxy disk and bulge ) of total mass of @xmath4 is assumed to be distributed among @xmath5 particles fragments .
the single particle density profile is also assumed to be a plummer one .
its mass is taken to be @xmath6 and radius @xmath7 kpc .
initially all particles are smoothly placed inside the sphere of radius @xmath8 kpc and are involved into the hubble flow ( @xmath9 km / s / mpc ) and solid body rotation around @xmath10 axis .
the initial motion of this system is described as : @xmath11 + h_{0}\cdot { \bf r } + { \bf dv}(x , y , z),\ ] ] where @xmath12 is an angular velocity of the rotating sphere , @xmath13 and @xmath14 the components @xmath15 , @xmath16 , @xmath17 of the random velocity @xmath18 vector are assumed to be initially randomly distributed within an interval @xmath19 km / s .
the dynamical evolution of baryonic matter fragments which are subjected to gravitational influences of dm baryonic halo and interfragment interactions is followed by means of effective n
body integrator with individual time step .
the dynamics of such n body system is described by following equations : @xmath20 the acceleration of @xmath21 - th particle @xmath22 is defined as a sum of three components .
@xmath23 where the first term @xmath24 accounts for gravitational interactions between fragments .
the second one @xmath25 is defined as an external gravitational acceleration caused by the dm and baryonic halo .
the last term @xmath26 corresponds to the viscous decceleration of fragment when passing through the baryonic halo .
the gravitational interaction between fragments is defined as the interaction of @xmath27 plummer profile elements : @xmath28 here @xmath29 and @xmath30 .
accounting for that the halo dm and baryonic components are also plummer spheres the second term becomes : @xmath31 the form of the last term @xmath26 will be discussed in the next subsection .
the characteristic time step @xmath32 in the integration procedure for each particle is defined as : @xmath33 \label{eq : def_dt}\ ] ] where the @xmath34 and the @xmath35 . here
the @xmath36 is a numerical parameter equal to @xmath37 that provides a nice momentum and energy conservation over the integration interval of about @xmath38 gyr .
for example , in the conservative case ( when viscosity of the system is set equal to @xmath39 ) , the final total error in the energy equation is less than @xmath40 .
the viscosity term @xmath26 is artificially introduced into the model so as to match the results of more sophisticated sph approach on dynamical evolution of disk galaxies .
the best fitness of results of this simplified approach with sph modelling data ( see @xcite ) is achived when the momentum exchange between the baryonic halo and moving particles is modelled by the following expression : @xmath41 here @xmath42 and @xmath43 are numerical parameters set equal to @xmath44 @xmath45 and @xmath46 kpc .
the vector @xmath47 is a particle velocity vector .
it is to be noted that single particle is assumed to have a total mass which does nt change with time and is defined as sum @xmath48 .
but masses of its gas and star components @xmath49 and @xmath50 are variable values and are defined by the temporal evolutionary status of the given fragment .
initially @xmath51 .
the results of fitness show that for @xmath10 component of viscosity term @xmath52 . in the galaxy plane where it is necessary to account for baryonic halo and baryonic fragments partial corrotation the dynamical friction is decreased and for @xmath53 component of viscosity term @xmath54 reduced to the value @xmath55 .
in the frame of the multifragmented model the definition of local gas density is introduced in the sph manner , e.g. local gas density depends on the total mass of matter contained in the sphere of radius @xmath56 around the @xmath21 th particle . for each @xmath21 th particle the value of its smoothing radius @xmath57 is chosed ( using the quicksort algorithm ) requiring that the volume of such radius compraises @xmath58 nearest particles ( i.e. @xmath59 1 % of total number of particles @xmath27 ) .
therefore , the total mass @xmath60 and density of gas @xmath61 inside this sphere are defined as @xmath62 where @xmath63 is defined as @xmath64 @xmath65 @xmath66
a forming disk galaxy is modelled as a system of interacting fragments ( named as particles ) embedded into the extended halo .
each particle is composed of gas and stellar components and its total mass is defined as @xmath67 .
initially all particles are purelly gaseous and , therefore , initially @xmath68 . to follow a particle star formation ( sf ) activity a special timemark @xmath69
is introduced which initially is set equal to @xmath70 .
the particles eligible to star formation events are chosen as particles which still have a sufficient amount of the gas component and their densities exceed some critical value @xmath71 during some fixed time interval @xmath72 ( of order of free fall time ) : @xmath73 here @xmath74 myr , and @xmath75 @xmath45 ( this last value is not crucial and is only limiting one ) .
if the particle was subjected to sf activity the parameter @xmath76 is set equal to @xmath77 , , and @xmath78 and @xmath79 are redefined as @xmath80 here @xmath81 is a sf efficiency which is defined as @xmath82 therefore , @xmath83 @xmath84 to match available observational data on star formation efficiency ( see @xcite ) parameters @xmath85 and @xmath86 are set equal to @xmath87 and @xmath88 . the chemical evolution of each separate fragment is treated in the frame of one zone close box model with instantaneous recycling . in the frame of this approach for the returned fraction of gas from evolved stars
is taken a standard value @xmath89 . following the instantenuous recycling approximation it is assumed that after each sf and sn explosions the heavy element enriched gas is returned to the system and mixed with old ( heavy element deficient ) gas .
after each act of sf and sn explosions the value of heavy element abundances of gas in particle is upgraded according to : @xmath90 the value @xmath91 is used as an average value for all values of @xmath92 ( see @xcite ) .
initially @xmath93 in all particles .
the proposed simple model provides the self - consistent picture of the process of galaxy formation , its dynamical and chemical evolution is in agreement with the results of more sophisticated approaches ( see @xcite ) . *
the rapidly rotating protogalaxy finally formed a three - component system resembling a typical spiral galaxies : a thin disk and spheroidal component made of gas and/or stars and a dark matter halo .
+ fig.1 and fig.2 show respectively the star formation rate and the galaxy total stellar mass as a function of time .
fig.3 shows the cylindrical distribution of stellar ( upper curve ) and gaseous ( lower one ) components of the final model disk galaxy as a function of a distance from the galaxy center in the galactic plane .
+ the total star formation rate ( sfr ) is a succesion of short bursts which does nt exceed @xmath94/yr . during first @xmath95 gyr of evolution only about @xmath96 of total galaxy mass is transforms into the stars
. the sfr gradually decreases , during the further evolution , to the value of about @xmath97/yr typical for our own galaxy .
the final total stellar and gas mass of the model galaxy disk are about @xmath98 and @xmath99 .
all these data as well as surface densities distributions of stellar and gaseous components ( see fig.4 ) are in nice agreement with present date observational data ( @xcite ) .
* the disk component posesses a typical spiral galaxy rotation curve and the distribution of radial and @xmath100 - th velocities of baryonic particles clearly show the presence of the central bulge ( see fig.6 and fig.7 ) . * the metallicities and the global metallicity gradient resemble distributions observed in our own galaxy ( fig.5 ) .
the averaged observed value of global metallicity @xmath101 ( see @xcite ) is shown in this fig . as a solid line .
* acknowledgements : * peter berczik would like to acknowledge the american astronomical society for financial support of this work under international small research grant .
figure 1 : the variation of total star formation rate of forming disk galaxy with time .
figure 4 : the surface density radial distributions of stellar and gaseous components ( stellar component is shown by filled stars , gaseous one by astericks ) . theoretical distributions ( not scaled ) of surface density for radial exponential scale lengthes @xmath102 and @xmath103 kpc are shown below by lines . | the evolving galaxy is considered as a system of baryonic fragments embedded into the static dark nonbaryonic ( dh ) and baryonic ( bh ) halo and subjected to gravitational and viscous interactions .
although the chemical evolution of each separate fragment is treated in the frame of one zone close box model with instantaneous recycling , its star formation ( sf ) activity is a function of mean local gas density and , therefore , is strongly influenced by other interacting fragments . in spite of its simplicity
this model provides a realistic description of the process of galaxy formation and evolution over the hubble timescale . |
the waveguide structure fabricated from the two closely placed waveguides is of common use in the fiber and integrated optics .
coupling between the waveguides is due to tunnel penetration of light from one waveguide into another waveguide @xcite .
this coupler preserves direction of light propagation , and for this reason it is named a directed coupler .
it was found @xcite that the steady state pair of electromagnetic pulses can exist in the extended directed coupler or twin - core fibers @xcite .
sometimes it termes soliton .
there is a large body of publications devoted to investigation of the soliton generation and its propagation , see for example @xcite .
recent progress in nanofabrication has lead to the design of new materials with highly unusual optical properties @xcite , of which__negative refraction _
_ is an example .
negative refraction occurs in media in which the wave vector of the electromagnetic wave is antiparallel to the poynting vector @xcite . in the particular case
when the real parts of the dielectric permittivity and magnetic permeability in the medium simultaneously take on negative values in some frequency range , the property of negative refraction will appear .
the existence of such media was demonstrated experimentally first in the microwave and then in the near - infrared ranges @xcite .
negative refraction can be realized for the media with spatial dispersion @xcite which is large enough and in the photonic crystals @xcite .
recently bulk metamaterials that show negative refraction for all incident angles in the visible region were presented @xcite .
the unusual properties of negative refractive ( nr ) index materials reveal themselves most prominently when the refractive index of the same medium can be positive in one spectral region and negative in another @xcite .
new wave propagation phenomena can also be expected when a wave passes through , or is localized near , an interface between such a material and a conventional dielectric , i.e. , positive refractive ( pr ) index material @xcite .
we can refer to these cases as _ negative - positive refraction _ ones .
the intriguing example of the negative - positive refraction medium is the coupler , where one of the waveguides is fabricated from a material with a negative refractive index .
this device acts as a ( distributed ) mirror .
the radiation entering one waveguide leaves the device through the other waveguide at the same end but in the opposite direction .
for this reason , this device can be called a _
oppositely - directed coupler _ ( odc ) .
the odc is known to support the propagation of linear waves with a gap in their spectrum ( forbidden zone ) .
it has been shown recently @xcite that due to this gap the _ nonlinear _ oppositely directed coupler is bistable .
bistability results from the multi - valued dependence of the transmission coefficient on the input - wave power .
it is noteworthy that this effect has no analogy in conventional directed couplers consisting of uniform waveguides without a mirror - based feedback mechanism .
it was found @xcite that if the waveguide nonlinear optical properties are characterized by the third order susceptibility , a coupled pair of the steady state pulses ( each one is localized in its own waveguide ) can exist in odc .
the fabrication of the transparent nonlinear nr waveguide is a complicated and still unresolved problem .
the linear nr waveguide could be more suitable for realization .
however there is question is it possible a quasisolitonic regime for the waves in such a _ antisymmetric _ coupler consisting of a nonlinear conventional dielectric waveguide coupled with a linear nr waveguide . in this paper
we consider the extended asymmetric oppositely - directed coupler ( aodc ) .
we found the steady state solutions of the equations which describe the waves in the aodc .
the numerical simulation of these steady state pulses formation and the collisions between the quasisolitons is performed .
we found that the pulses are very robust against perturbations .
it allows a definite conclusion that the steady state pulses are a new kind of gap solitons .
it should be noted that this solitary wave exists in the medium without grating . the influence of linear losses in the nr waveguide on the quasisoliton formation discussed shortly .
we consider the waveguide structure in which one of the channels is made from nonlinear material features by the positive refraction , and the nr channel is composed of linear material .
the system of the equations based on connected wave approximation takes the form @xmath0 the coefficient @xmath1 is the effective non - linear susceptibility of the first channel . @xmath2 and @xmath3 are the coupling constants .
linear properties of the pr channel defined by dielectric permittivity @xmath4 and magnetic permeability @xmath5 .
these values are assumed be real , that corresponds to the lossless materials .
slowly varying envelope of the electric field @xmath6 in @xmath7-th channel is written as @xmath8 , @xmath9 is the group - velocity of the wave in @xmath7-th channel , @xmath10 . here
we assume that the group - velocity dispersion can be neglected . in the following
we also assume that the synchronism condition is satisfied : @xmath11 .
it is convenient to use new normalized variable @xmath12 in the normalized variables the ultimate system of equations ( [ eq1 m ] ) for the aodc reads as @xmath13 the parameter of nonlinearity is @xmath14 let us consider the linear wave limit of the aods equations
. from ( [ eq2 ] ) one can get @xmath15 if we take the fourier transformation @xmath16 the equations ( [ eq2lin ] ) result in the following linear system of equation @xmath17 this system of equations has the nonzero solution only if the corresponding determinant @xmath18 is equal to zero .
that leads to the dispersion relation @xmath19 thus the spectrum of the linear waves has the gap @xmath20 .
this gap is characteristic feature for a distributed mirror @xcite .
hence , the aodc in linear wave limit acts as a mirror . in following consideration
it is suitable to take the real variables form of equations ( [ eq2 ] ) . by using the real variables
@xmath21 one obtains @xmath22 @xmath23 @xmath24 where @xmath25 . from the amplitude equation
it follows that @xmath26 hence , @xmath27 in the case of solitary waves , for which the electromagnetic fields vanish at infinity are considered , the right part of ( [ eq4 ] ) is equal to zero .
it leads to integral of motion @xmath28 it is the modified manley - rowe relation .
as usually , in the case of the quadratic nonlinear pr media manley - rowe relation looks like @xmath29 the difference of the expression ( [ intmot ] ) from conventional manley - rowe relation is explained by the fact that the energy flux for the waves in one waveguide opposite in direction to energy flux of other waveguide , while their wave vectors are approximately the same .
this flux pattern is an inherent feature of the negative - positive refraction media .
to consider the solitary steady state waves in aodc by similar way as in @xcite we have start from the equations ( [ eq3 ] ) .
suppose that solutions of these equations are depend only on single variable @xmath30 with free parameter @xmath31 .
suppose @xmath32 and @xmath33 .
the system of the equations ( [ eq3 ] ) takes the following form @xmath34 @xmath35 where @xmath36 we can also write an equation for the phase difference @xmath37 the phase equations can be used to get one equation for the phase difference .
finally , the total system of equations reads @xmath38 we are looking for a solution in a form of the solitary wave , it corresponds with the following boundary condition @xmath39 from the equation ( [ eq7a ] ) and ( [ eq7b ] ) it follows @xmath40 , or @xmath41 , where @xmath42 .
hence , the system of equations ( [ eq7a])-([eq7c ] ) is reduced to following pare of equations @xmath43 multiplying the last equation by @xmath44 and taking into account the equation ( [ eq8a ] ) , we get the second integral of motion @xmath45 due to the boundary condition , the value of this integral is equal to zero .
as we are looking for a non zero solution of ( [ eq8a ] ) and ( [ eq8b ] ) , one can write @xmath46 substitution of ( [ eq10 ] ) into ( [ eq8a ] ) leads to @xmath47 choosing the variable @xmath48 reduces this expression into equation @xmath49 which has the following solution @xmath50 thus , the solution of ( [ eq8a ] ) and ( [ eq8b ] ) is @xmath51 by using the phase s equations and ( [ eq11 ] ) one can write @xmath52 @xmath53 that yields @xmath54 the phases @xmath55 should be chosen in such a way that ( [ eq10 ] ) is satisfied .
as the solitary wave s amplitude tends to zero @xmath56 , the relation ( [ eq10 ] ) at infinity reduces to @xmath57 in the numerical simulation we choose @xmath58 and @xmath59 .
the amplitudes @xmath60 are defined by the following expressions @xmath61 the expressions ( [ eq13a ] ) and ( [ eq13b ] ) describe the steady state solitary wave propagating in aodc under consideration .
this wave looks like a gap soliton propagating in the nonlinear bragg grating , however there are no any periodic structures .
gap in the linear wave spectrum ( [ disprel ] ) is due to the flux pattern in this negative - positive refraction media .
some remarks are worthy to be made about the free parameter @xmath31 .
the negative value of the parameter @xmath31 corresponds to the solitary wave propagating in the direction of the axis @xmath62 . the solitary wave characterized by positive value of the parameter @xmath31 propagates in the opposite direction .
the large amplitudes ( more powerful solitary waves ) correspond to large positive values of the parameter @xmath31 . for the negative values of the parameter
@xmath31 the quasisolitons with smaller amplitudes have smaller values of the parameter @xmath31 however the absolute value of the velocity determined by the parameter @xmath31 is larger for the less powerful solitary waves
. we will refer to the fig.[fig1 ] later in discussion the influence of the linear losses at the quasisoliton .
it is common knowledge that the completely integrable evolution nonlinear equations have a special solutions describing elastic interaction of solitary waves @xcite .
these waves are named solitons .
it is our opinion that the system of equations ( [ eq2 ] ) does not belong to the class of completely integrable equations .
hence the solution of these equations does not represent true soliton .
we will denote them as quasisolitons . to investigate interaction between the steady state solitary waves ( [ eq13a ] ) and ( [ eq13b ] )
the numerical simulation was pursued . numerical solution of the system of equations ( [ eq2 ] ) is performed using the scheme of the finite differences .
the conservation of the first integral of motion ( [ intmot ] ) is used to control the computational error . in the consequent an integration step over the evolution variable @xmath63
is set to @xmath64 .
the nonlinearity is set to @xmath65 .
( the smaller nonlinearities correspond to the quasisolitons with the same profiles but the smaller amplitudes ( defined by the parameter @xmath66 which is linearly proportional to the nonlinearity coefficient . )
it was found that the results of the collision depend on the relative velocity of the pulses .
fig.[m05m09 ] illustrates collision between quasisolitons characterized by @xmath67 and @xmath68 .
the relative interaction velocity is @xmath69 .
the quasisolitons drop some radiation after interaction and weakly radiating wave could be noticed in the left panel of the picture as a result of collision .
velocities and the amplitudes of the resulting quasisolitons are slightly different from the parameters of the quasisolitons before the collision due to loss of some part of energy .
decrease the difference between velocities of colliding pulses results in long - range interaction between pulses .
the interaction distance depends on the relation of the velocities . in the fig.[m05m07 ] is shown interaction of two quasisolitons characterized by @xmath70 and @xmath68 .
the relative interaction velocity is @xmath71 .
the quasisolitons exchange energy at the distance and form a transitory bound state .
it was mentioned before that the quasisolitions with the same absolute values of velocity have different amplitudes . to investigate the robustness of the quasisoliton depending on its energy , at the fig.[m05m07 ] we present results of modeling the collision between two steady - state solitary waves with the same absolute values of the velocity @xmath72 and @xmath70 .
a more energetic quasisoliton with @xmath72 remains unchanged after collision and the less energetic quasisoliton with @xmath70 loose some radiation and changes its trajectory . thus , the collision of two steady state pulses with different velocities has shown significant robustness .
small amplitude radiation appearing after collision attests that the quasisoliton eventually be disappeared .
for the current state of fabrication technology the losses in the real nr materials are considerable .
high value of losses renders steady state pulse propagation impossible .
nevertheless , the role of small losses would be considered . to check the influence of linear losses in the nr channel on the solitary wave formation and propagation , the model equations ( [ eq2 ] ) were modified by including the additional term in the equation for a nr channel @xmath73 the fig.[07m07dec ] illustrates the evolution of same pulses as in the fig.[07m07 ] placed at the boundaries however the linear losses in the nr channel are taken into consideration @xmath74 . comparing these two pictures
one may conclude that even small losses affect considerably the propagation properties of the solitary waves . propagating in the dissipative medium these pulses loose energy and the pulse with positive
@xmath31 slows down to sustain the steady state regime .
the pulse with negative @xmath31 slightly accelerates to support the quasi - steady regime of propagation ( notice a fig.[fig1 ] ) .
we considered the nonlinear solitary waves propagating in the nonlinear opposite - directed coupler .
one of its components is a nonlinear waveguide made from the material with positive index of refraction .
another channel is fabricated from linear dielectric material with negative refraction index .
for the system of coupled waves the analytical solution for the electromagnetic waves in the form of steady state solitary waves is found .
it is interesting to notice that the wave in the nonlinear channel affects the wave propagation in the neighboring linear channel .
this influence results in coupling and a steady propagation of the solitary waves in both waveguides , the wave in the nonlinear waveguide draws the wave in the linear nr waveguide .
it was shown that the result of interaction of the quasisolitons depends on their velocities .
provided that relative velocity of the two colliding quasisolitons is large , the pulses collide almost like the solitions . a weak radiation or a weak pulse emerges as a result of collision .
at small relative velocities of the quasisolitons the interaction between two colliding solitary waves could results in a strong energy exchange between colliding solitary waves and formation of the temporarily coupled state of the interacting quasisolitons .
this phenomena is known for the case of nonlinear directed coupling . in the case under consideration the formation of the coupled state demands for an additional investigation .
it could be possible that a long - living coupled state of two solitary waves is exists there .
the influence of the linear losses in the nr waveguide on the existence of the solitary waves was studied .
the wave loses its energy as it propagates through the dissipative medium . in order to compensate the losses the entering pulse in a form of the solitary wave characterized by positive value of parameter @xmath31 (
its propagation direction is opposite to the coordinate axis @xmath62 ) permanently transforms into another solitary wave , less energetic , having smaller value of the parameter @xmath31 .
it results in slowing down the wave in the lossy medium .
after that the wave stops in the waveguide and changes the direction of its propagation .
oppositely the entering pulse in a form of the solitary wave characterized by negative parameter @xmath31 ( its propagation direction is coincides to the coordinate axis @xmath62 ) tends to increase its velocity ( i.e. to decrease parameter @xmath31 which is negative ) to sustain the steady - state regime of propagation in presence of the losses .
we are pleased to thank our colleagues prof .
s.o . elyutin and prof .
i.r . gabitov for useful discussions .
maimistov appreciates the support and hospitality of the department of mathematics in the university of arizona during his work under this paper .
the research of a.i .
maimistov and s.s .
ozhenko was partially supported by rfbr(n 06 - 02 - 16406 ) . | we consider the electromagnetic waves propagating in the system of coupled waveguides .
one of the system components is a standard waveguide fabricated from nonlinear medium having positive refraction and another component is a waveguide produced from an artificial material having negative refraction .
the metamaterial constituting the second waveguide has linear characteristics and a wave propagating in the waveguide of this type propagates in the direction opposite to direction of energy flux .
it is found that the coupled nonlinear solitary waves propagating both in the same direction are exist in this oppositely - directed coupler due to linear coupling between nonlinear positive refractive waveguide and linear negative refractive waveguide . the corresponding analytical solution is found and it is used for numerical simulation to illustrate that the results of the solitary wave collisions are sensible to the relative velocity of the colliding solitary waves . |
the effect of a treatment on an outcome is often the main parameter of interest in many scientific fields .
the standard approach used to estimate it is the so called average treatment effect ( ate ) , the difference between the expected values of the response s distributions under the two treatment regimes .
while intuitive and useful in many situations , it suffers from some limitations ; in particular , it becomes highly biased when the response is skewed .
a further drawback of the ate is its coarseness as a summary of the distance between the expected value of the response s distributions under the two treatments .
it is a matter of fact indeed that the effect of the treatment on the outcome often varies as we move from the lower to the upper tail of the outcome s distribution .
this limitation of the ate has been addressed in the literature by introducing the so called quantile treatment effect ( qte ) , the difference between the response s distribution quantiles under the two treatments @xcite . in this paper
we propose a more general measure of the effect of a binary treatment on a continuous outcome .
we call it the _ generalized quantile treatment effect _ ( gqte ) , defined as @xmath1 where @xmath2 and @xmath3 represent the quantile functions of the outcome under the two treatment conditions and @xmath4 is an arbitrary but known function of the quantiles . for example , if @xmath5 is chosen to be the identity function , then the gqte simplifies to the qte , while if @xmath4 is the integral over the percentile @xmath6 , the gqte becomes equivalent to the ate .
the gqte is a new parameter which generalizes the existing approaches for estimating a treatment effect . to estimate and formulate inferences about the gqte we propose a bayesian approach that can accommodate both symmetric and skewed outcomes , as well as situations where the sample size under a treatment condition ( cases ) is much smaller than the sample size under the other treatment condition ( controls ) . in particular , we assume @xmath7 where @xmath8 is a monotone function and @xmath9 is assumed to be smooth .
in other words , we assume that the transformed quantile ratio is a smooth function of the percentile @xmath6 .
the idea of smoothly modeling the ratio of the quantiles has been first introduced by @xcite , who exploited it by proposing a nonparametric estimator of the mean difference between two populations . here
we generalize their approach by permitting the comparison of any characteristic of the outcome s distributions under the two treatments .
an important theoretical contribution of this paper is the derivation of a closed form expression for the model likelihood .
we show that it is possible to obtain an analytically tractable form for the @xmath10 density ( the controls ) without explicitly specifying a model for it .
clearly the likelihood is needed to carry out the bayesian estimation but in principle it could be employed for classic likelihood procedures as well .
moreover , our proposed approach allows one to borrow strength from one sample to the other , thus improving efficiency in the estimation of the quantiles @xcite . as an illustration
, we apply our method to the comparison of the single hospitalization medical costs distribution between subjects with and without smoking attributable diseases .
the data set we use is the national medical expenditures survey ( nmes ) supplemented by the adult self - administered questionnaire household survey .
the paper is organized as follows . in section [ gqte ]
we define the new parameter @xmath11 and illustrate some quantile - based measures that will be used in the paper . in section
[ estimation ] we provide details of the estimation approach together with some special cases .
we then present the results of a simulation study in section [ simstudy ] through which we conclude that under a broad set of conditions our approach performs better than other flexible methods for comparing two distributions . in section [ data_analysis ]
we illustrate the results of the data analysis on the nmes data set .
section [ discussion ] concludes the paper with a discussion and some final remarks .
consider two positive continuous random variables @xmath12 and @xmath13 with quantile functions @xmath14 and @xmath15 , where @xmath16 for @xmath17 and @xmath18 . to compare @xmath19 and @xmath20 as flexibly as possible
we introduce the generalized quantile treatment effect , which is defined as @xmath21 where @xmath4 is a known function of the quantiles .
notice that no a priori assumptions are made about the admissible functions @xmath5 , thus potentially any function of the quantiles can be used . therefore , the gqte provides a general approach to compare the response s distributions under the two treatments .
more precisely , by properly choosing the function @xmath4 , we can recover any specific characteristic of the outcome s distributions @xmath22 and @xmath23 and , through , their difference .
the simplest case arises when @xmath24 . in this case
the gqte simplifies to @xmath25 the so called ( unconditional ) qte @xcite , sometimes also named the percentile - specific effect between two populations ( see for example * ? ? ?
* ; * ? ? ?
a second example is obtained by choosing @xmath26 , which produces @xmath27 the extensively used ate ( see for example * ? ? ?
* chapter 21 ) .
these examples illustrate how the gqte reduces to the two most used parameters of interest for estimating a treatment effect , the ate and qte .
however , the gqte can provide a variety of other useful measures . in appendix 1
we illustrate some other interesting cases that usually are not taken into consideration in the literature .
in this section we illustrate the procedure we developed for estimating the gqte .
our proposed approach is sufficiently general that it can be used for any choice of @xmath4 .
we assume that @xmath28 , where @xmath22 is a given probability distribution depending upon a vector of unknown parameters @xmath29 .
for example , in the application presented in section [ data_analysis ] we choose @xmath22 as a mixture distribution .
to borrow information from one distribution to the other , we assume that the transformed quantile ratio is a smooth function of the percentiles with @xmath30 degrees of freedom , that is @xmath31 the function @xmath32 is assumed to be monotone differentiable . it represents a kind of link function and
it is used to transform the quantile ratio to account for the potential skewness of the @xmath22 and @xmath23 .
the typical choice for skewed data is @xmath33 , while for symmetric data distributions the identity function is the most reasonable option . for the sake of simplicity
, we henceforth indicate the smooth function @xmath34 with reference to the corresponding design matrix @xmath35 , so that it can be written as @xmath36 , where @xmath37 is a vector of unknown parameters .
more explicitly , we assume that @xmath38 , where @xmath39 are orthonormal basis functions with @xmath40 .
the number of degrees of freedom @xmath30 is a further parameter that has either to be chosen or estimated from the data . in subsection [ dfselection ]
we propose a simple approach for eliciting it .
the basis functions are usually either splines or polynomials .
the main justification for assuming is that it allows one to borrow information from both the response s distributions under the two treatment conditions when we estimate the gqte @xmath41 .
assumption , in fact , implies @xmath42\,,\ ] ] and also @xmath43\right\}^{-1}\,,\ ] ] which , once substituted in , return @xmath44 \big ) - g\big ( q_1(p ) \left\{h^{-1}\left[x(p,\lambda)\ , { \boldsymbol}{\beta}\right]\right\}^{-1 } \big).\ ] ] for the special case where @xmath24 and @xmath45 , @xcite have shown that under assumption it is possible to obtain a more efficient estimator of @xmath46 than the sample mean difference and the maximum likelihood estimator assuming that @xmath47 and @xmath10 are both log - normal .
notice that , since the main interest in the paper resides in the estimation of @xmath48 , for which only @xmath37 is required , @xmath29 is treated as nuisance ( see subsection [ gqte_inference ] for further details ) .
in this paper we propose a bayesian approach for estimating @xmath41 for any choice of @xmath4 and @xmath32 .
an interesting feature of our estimation procedure for @xmath48 , is that we only need to specify the distribution function for @xmath47 .
the specification of @xmath19 together with the relationship automatically determines a distributional assumption for @xmath13 .
we refer to the distribution of @xmath10 induced by @xmath22 and assumption as @xmath49 . as a last remark for this section , we want to highlight the difference between the function @xmath4 , introduced in the previous section , and @xmath32 , defined above in .
they should not be confused because they have distinct roles : the former identifies the response s characteristic we want to estimate for assessing the treatment effect , while the latter has been introduced as a mechanism to attenuate the possible skewness present in the data .
the steps involved in our estimation approach are summarized as follows : 1 .
choose a ( possibly flexible ) density @xmath50 for @xmath47 , a smoothing function @xmath34 ( usually a spline or a polynomial ) and a value for @xmath30 ; 2 .
from derive the density function of @xmath10 , that we denote as @xmath51 .
note that , as proved by theorem [ theorem1 ] below , this density will depend on the model parameter @xmath37 as well as on the parameter @xmath29 through the @xmath47 density .
3 . calculate the joint likelihood @xmath52 to use for finding the posterior distribution of @xmath53 in a markov chain monte carlo ( mcmc ) algorithm .
4 . obtain the posterior distribution of any special case of the gqte . the critical step in this sequence
is represented by the calculation of the likelihood , which we now describe .
consider two i.i.d .
samples @xmath54 and @xmath55 drawn independently from the two populations @xmath56 and @xmath49 .
we refer to the former as the cases ( or the treated ) and to the latter as the controls ( or the untreated ) .
we assume that these distribution functions have densities @xmath57 and @xmath58 respectively .
the likelihood function for our model is then given by @xmath59 since we did nt state any specific distributional assumption for @xmath10 , in principle we could not calculate the likelihood because we do nt have any expression for @xmath60 .
two strategies are possible here . given the @xmath61 specification , one possibility is to find an expression for @xmath62 , then map it through equation to find a corresponding expression for @xmath63 , invert it to determine @xmath23 , and finally differentiate the result to get @xmath60 .
apart from simple situations , usually these steps ( i.e. integration , inversion and differentiation ) need to be performed numerically .
a second possibility is to replace the @xmath10 density in the likelihood with its correspondent _ density quantile function _
, @xmath64 ( see * ? ? ?
* ) , for which the next theorem provides a closed form expression .
the proof of the theorem and two additional corollaries are available in appendix 2 , while in the next subsection we provide some further explanation on how to compute @xmath60 .
[ theorem1 ] let @xmath28 , with @xmath22 having density function @xmath65 , and assume that holds .
if , for every @xmath17 , the vector @xmath66 satisfies the constraint @xmath67 \right\ } \leq\frac{1}{f_1\left(q_1(p)|{\boldsymbol}{\eta}\right)q_1(p)}\,,\ ] ] the density quantile function @xmath68 for @xmath10 is @xmath69|{\boldsymbol}{\eta } \right ) h^{-1}\left[x(p,\lambda)\,{\boldsymbol}{\beta}\right]}{1-f_1\left ( q_2(p)\,h^{-1}\left[x(p,\lambda)\,{\boldsymbol}{\beta}\right]|{\boldsymbol}{\eta } \right)\,x'(p,\lambda)\,{\boldsymbol}{\beta}\,q_2(p)\left\{\frac{d}{d\left ( x(p,\lambda)\,{\boldsymbol}{\beta}\right)}h^{-1}\left[x(p,\lambda)\,{\boldsymbol}{\beta}\right]\right\ } } \,.\ ] ] the function @xmath68 is a properly defined density . note that @xmath70 correctly depends upon both the model parameter @xmath37 and the @xmath12 parameter @xmath29 through the @xmath71 density .
the motivation for the constraint comes from the need to guarantee that @xmath60 is a non - negative function . as a further remark
, we observe that the term in the likelihood involving @xmath72 depends upon the observations @xmath73 through the unknown quantile function values @xmath74 .
a computational drawback of our proposal is that the `` true '' values of the percentiles @xmath75 , i.e. those generated under the assumed model for @xmath10 , should be used in the calculation of the likelihood .
unfortunately , these are not available , because the cumulative distribution function @xmath23 is not given explicitly and we can not find the @xmath75 corresponding to the observed data @xmath73 as @xmath76 .
the approach we recommend to bypass this issue is to approximate the @xmath75 using the procedure described in @xcite , which we summarize as follows : 1 . denoting with @xmath77
the ordered observed values for @xmath10 , we look for the corresponding set of ordered @xmath78 such that @xmath79 , where @xmath80 is an estimate of @xmath3 based on the current values of the parameters @xmath37 and @xmath29 ( i.e. the values from the current mcmc draw ) .
more specifically , we find the @xmath78 using the following procedure : suppose @xmath81 is the current estimate of @xmath6 for a given @xmath82 value .
then , for a value of @xmath6 close to @xmath81 , @xmath83 can be approximated using the following taylor series expansion @xmath84 which , solving for @xmath6 , gives @xmath85 where @xmath86 is the quantile density function corresponding to @xmath87 and where we used the fact that @xmath88 . as a starting point for @xmath78
we use @xmath89 , @xmath90 .
equation is used in an iterative fashion till the given value of @xmath83 differs from @xmath82 by less than some chosen small amount ( we use @xmath91 ) .
once the values of @xmath75 are available , we compute the quantities @xmath92 using equation .
the critical issue in this step is the calculation of the derivative @xmath93 . in the cases we consider here ( i.e. either a polynomial or a spline basis ) , the derivative is available in closed form and
so no further numerical approximation is needed .
strictly speaking , the calculation of @xmath60 provided by the procedure we just described is not exact but involves a numerical approximation .
we performed a detailed analysis on the goodness of this approximation and we found that the actual and approximated @xmath60 values ( and hence the overall likelihood ) were indistinguishable .
we present now some special cases where an appropriate choice of the design matrix @xmath35 allows to recover an exact expression for @xmath68 belonging to a known distribution family .
the proofs of these special cases are provided in appendix 3 .
_ case 1 : @xmath12 is uniform and @xmath94 . _ in this case we assume that @xmath95 $ ] and choose @xmath96 .
then @xmath97 and from it follows that @xmath98}\{q_2(p)\}\,,\ ] ] which corresponds to the density quantile function of a uniform random variable with parameter @xmath99 , where @xmath100 denotes the indicator function taking value @xmath101 if @xmath102 and @xmath103 otherwise .
note that it correctly depends both upon the parameter @xmath104 and the @xmath12 parameter @xmath105 .
_ case 2 : @xmath12 is log - normal and @xmath106 $ ] .
_ we now assume @xmath107 and fix @xmath45 .
it follows that @xmath108 , where @xmath109 is the quantile function of a standard normal random variable .
then by we get @xmath110 ^ 2}{2(\sigma_1-\beta _ 1)^2}\right\}\,,\ ] ] which is the density quantile function of a @xmath111 random variable with @xmath112 and @xmath113 .
note that the density quantile function of @xmath13 correctly depends both upon the parameters @xmath114 and the @xmath12 parameters @xmath115 . in this case
the constraint simply requires that @xmath116 , for every @xmath117 .
_ case 3 : @xmath12 is pareto and @xmath118 $ ] . _
suppose @xmath119 and choose @xmath120 .
in this case @xmath121 and by we get @xmath122 which represents the density quantile function of a @xmath123 random variable with @xmath124 and @xmath125 .
the density quantile function of @xmath13 correctly depends both upon the parameters @xmath126 and the @xmath12 parameters @xmath127 . in this case
the constraint requires that @xmath128 , for every @xmath117 . the parameters @xmath37 and @xmath29 are assumed to be a priori independent , that is @xmath129 where @xmath130 and @xmath131 are the prior hyperparameters for @xmath37 and @xmath29 respectively .
for @xmath132 we use a a multivariate normal distribution with mean equal to the ordinary least squares ( ols ) estimate of @xmath37 based on the model @xmath133 where @xmath134 , @xmath135 and variance - covariance matrix equal to @xmath136 , where @xmath137 is normally fixed at a high value to induce a weakly informative prior distribution for each @xmath138 and @xmath139 indicates the identity matrix with size @xmath140 . for the prior distribution for @xmath29 , the choice clearly depends upon the assumption made about @xmath22 , but we suggest to use conjugate priors . for an example see the application in section [ data_analysis ] .
the posterior distributions of @xmath37 and @xmath29 are obtained by an mcmc simulation . in particular
, we use an independent metropolis - hastings algorithm with blocking over @xmath37 and @xmath29 separately ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . as the proposal distribution for @xmath37
we use a @xmath140-dimensional @xmath141 distribution with mean and scale matrix chosen to match the @xmath142 ols estimate and variance from , and a small number of degrees of freedom , usually set to @xmath143 . as with the prior , the proposal distribution for @xmath29 depends upon the particular application under investigation ( see section [ data_analysis ] for an example ) . once
the @xmath142 parameters have been estimated and the convergence of the simulated chains has been assessed by conventional methods ( see * ? ? ? * , or ; * ? ? ?
* ) , we can obtain the posterior distribution of the gqte for any choice of @xmath4 as we now describe . for each iteration @xmath144 of the mcmc simulation ,
a value @xmath145 for @xmath37 is available . using expressions and
we can obtain @xmath146 and @xmath147 as @xmath148 \
, , \qquad i=1,\ldots , n_{2},\ ] ] and @xmath149 \right\ } ^{-1 } \ , , \qquad i=1,\ldots , n_{1},\ ] ] where the @xmath150 are the order statistics for sample @xmath151 , while @xmath152 , @xmath153 , for @xmath154 .
it then follows that the @xmath144-th iteration value for @xmath11 is given by @xmath155 where @xmath156 and @xmath147 are found by interpolating the estimated quantile functions @xmath157 and @xmath158 .
the estimate of @xmath41 is finally obtained through the rao - blackwellized estimator @xmath159 where @xmath160 is the total number of iterations .
since the whole posterior distribution of @xmath11 is available , standard inferential questions can be easily addressed in the usual ways .
as detailed in section [ gqte ] , many interesting special cases arise from the general definition of the gqte .
for example , if interest lies in estimating the qte defined in , @xmath4 corresponds to the identity function and expression becomes @xmath161 which can be evaluated for any value of @xmath162 .
if the focus is on the ate , defined in , then returns the estimator @xmath163 } \\ - \frac{1}{n_1}\sum^{n_1}_{i=1}{y_{1(i ) } \left\ { h^{-1}\left [ x\left(p_{1i},\lambda\right)\widehat{{\boldsymbol}{\beta}}^{(m ) } \right ] \right\}^{-1 } } \,.\end{gathered}\ ] ] appendix 4 contains the details for some other cases . as a final remark , note that an appealing feature of this approach is that the mcmc procedure needs to be run only once to compute the difference between any measure of the treatment effect of interest in the two groups .
the choice of the number of degrees of freedom @xmath30 to use in the procedure above is not trivial .
many approaches can be proposed , but to keep the computational burden manageable , we propose to elicit it by minimizing an empirical version of the @xmath165 discrepancy measure @xcite @xmath166 where @xmath167 denotes the unknown true @xmath10 density .
more precisely , we select @xmath30 using the following procedure : for each @xmath168 , where @xmath169 is the maximum admissible value for @xmath30 , we estimate @xmath170 , where @xmath171 is equal to the ols estimate given in and @xmath172 is estimated by using only the data @xmath173 . after replacing the true unknown density @xmath167 with a kernel density estimate of the data @xmath174 , the value of @xmath30 is chosen as that minimizing the value of over the set @xmath175 .
one drawback of this approach is that it tends to select high values of @xmath30 .
we provide further discussion about this issue in section [ discussion ] .
in this section we report the results of a simulation study we performed to compare the finite sample properties of our method with those of other flexible approaches.the simulation indicates that often the gqte procedure has lower mean squared error and a similar bias as other flexible methods for comparing two distributions . in particular , we contrast our proposal with the smooth quantile ratio estimation ( square ) approach presented in @xcite , and the probit stick - breaking process ( psbp ) proposed in @xcite .
the former is a frequentist semiparametric method while the latter is a bayesian nonparametric model .
we now provide some details about these two methodologies and a justification for using them .
in square it is assumed that the log quantile ratio is a smooth function of the percentile @xmath6 with @xmath30 degrees of freedom , that is @xmath176 the basic idea of smooth quantile ratio estimation is to replace the empirical quantiles with smoother versions obtained by smoothing the log - transformed ratio of the two quantile functions across percentiles .
square has been proposed by @xcite as an estimator of the mean difference between two populations with the advantage of providing substantially lower mean squared error and bias than the sample mean difference or the maximum likelihood estimator for log - normal populations . to estimate @xmath177 a @xmath178-fold cross - validation approach is suggested .
finite sample inference is performed by bootstrap but they also provide large sample results . the psbp is a general nonparametric bayesian model which has been proposed by @xcite for estimating the conditional distribution of a response variable given multiple predictors .
more specifically , the psbp is a prior for an uncountable collection of random distributions . like for the models belonging to the class of dependent dirichlet processes @xcite ,
the psbp main idea is to allow for dependence across a family of related distributions as a function of some covariates .
more explicitly , the psbp induce dependence in the weights of the stick - breaking representation @xcite by replacing the beta - distributed random variables with a probit model .
this simple change greatly enhances the flexibility of the model , thus providing an extremely interesting extension within the framework of dependent priors across families of probabilities measures .
we decided to compare the gqte with these two methods because they are both highly flexible and have been proved to perform well under a broad set of situations .
originally the simulation also included the anova dependent dirichlet process mixtures proposed by @xcite , but we decided not to report it here because it performed poorly as compared to the psbp model .
the reason for such inferior results resides in the definition of the model itself , which assumes the weights in the stick - breaking representation of the process to be fixed , i.e. the same for the response distributions under the two treatment conditions .
since square produces an estimate only for the mean difference between two populations , our simulation is restricted to this specific case .
we are aware that the results only provide a partial demonstration of the gqte advantages over the other methods , but we also need to stress that the ate is the most common measure used in practice for estimating the extent of a treatment effect .
our simulation framework is similar to that used in @xcite and includes five scenarios , which are described in table [ tabsim1 ] under the labels a to e. in scenarios a , b and c the @xmath10 distribution is assumed to be log - normal with parameters @xmath179 and @xmath180 which approximately correspond to the sample statistics for the medical expenditures of non - diseased subjects from the nmes data set . in scenario a , the @xmath47 distribution is also log - normal but with larger values of the parameters , namely @xmath181 and @xmath182 .
scenarios b and c use a different assumption for the @xmath47 distribution chosen to represent some reasonable shapes .
the next two scenarios , d and e , compare the performances of the different methods using real data . in particular , in scenario d the data are randomly drawn from the distributions of nonzero medicare expenditures for cases and controls from the nmes data set .
finally , scenario e assumes that both populations follow a gamma distribution with finite second moment .
.simulation study
sampling mechanisms under each simulation scenario . in scenario
d , @xmath183 are the empirical cumulative distribution functions of the nonzero medical expenditures for patients in the case and control groups from the nmes data set , and , in scenarios b and c , @xmath184 . moreover , in scenario b , @xmath185 , while in scenario c , @xmath186 . [ cols="^,^,^,^,^",options="header " , ] [ table_summary ] figures [ data_summary](b ) and ( d ) show the histograms and boxplots for the cubic root transformed data .
the need for such a transformation derives from the particular choice we make regarding the cases distribution ( see next subsection ) and is not a general requirement of our approach .
moreover , the use of this transformation does not alter in any way the results and the conclusions we draw , hence in the following we systematically refer to the transformed data . at any rate ,
note also that , even after the transformation , the outcome distributions still present heavy right tails .
this conclusion motivates the use of @xmath45 . in figure [ square_qqplot_cubic](a )
we report the q - q plot for the ( cubic root transformed ) nmes data .
we can identify a non - linear smooth relationship between the cases and controls medical expenditures .
panel ( b ) of the same picture , which shows the quantile ratio as a function of the percentile @xmath6 , confirms these findings . in this application
we assume that @xmath187 , a particular mixture of gamma distributions with density @xmath188 where the mixing occurs over the shape parameters and where @xmath189 , the number of components , is fixed a priori .
first introduced in @xcite , it has been explicitly developed as a model for right - skewed distributions and its parameterization allows to create a convenient and flexible method characterized by a single scale parameter for all the gamma components , plus the ordinary set of mixture weights .
we use conjugate priors @xmath190 and @xmath191 for the shared scale parameter and the mixture weights respectively .
the number of mixture components is fixed at @xmath192 , while the @xmath193 hyperparameters are set to @xmath194 and @xmath195 ( for more information on the elicitation of these priors see * ? ? ?
* section 2.3 ) .
the initial values of the metropolis - hastings algorithm are chosen as follows : the @xmath37 chain is started from its ols estimate , as discussed in , while for @xmath196 we first get a preliminary estimate ( with @xmath197 iterations ) using the approach described in @xcite , and then we fix their starting values to the corresponding estimated posterior averages .
we run the mcmc algorithm for @xmath198 iterations plus @xmath199 iterations as burn - in .
such a large number of iterations is necessary because the model , being quite complicated , has shown a slow convergence behavior of the chains .
the selection procedure described in subsection [ dfselection ] for the number of degrees of freedom suggested a value of @xmath30 equal to @xmath200 , which can be considered fairly satisfactory from a visual inspection of the scatterplot ( see figure [ square_qqplot_cubic](b ) ) .
the acceptance rates for the mcmc posterior simulations are relatively small , being around @xmath201 for @xmath37 , @xmath202 for @xmath193 and @xmath203 for @xmath204 . despite that
, we do not consider these results as problematic since the chain is moving in a high - dimensional space ( @xmath205 dimensions ) , which necessarily slows down the convergence process .
this is the main reason why we decide to run the simulation for a longer time .
however , the results of the analysis presented below indicate that convergence was attained .
we made other attempts with simpler ( but less flexible ) specifications of the @xmath12 distribution , which showed a more conventional behavior of the acceptance rates.=1 figure [ bsq_fit ] shows the fitted values for the estimated model .
the gray dots represent the quantile ratio for the transformed data as a function of the percentile @xmath6 .
the solid line illustrates the estimated posterior mean of the quantile ratio , while the dashed one represents its ols estimate ( the same line as in figure [ square_qqplot_cubic](b ) ) .
the shaded area gives the credible bands for the estimated posterior means , showing a fairly low amount of uncertainty around the estimates .
moreover , from the picture we can see that our model is less sensitive to extreme observations , especially in the right tails of the distributions.=1 credible intervals .
the dashed line corresponds to the ols fit for the same data , as described in . ] in figure [ est_f2 ] we report the estimated @xmath13 density .
it is possible to ascertain a quite good fit . in the display , together with the @xmath206 posterior mean , we put the corresponding @xmath207 credible bands and the histogram of the data
. density .
the solid line represents the estimated pointwise posterior means , while the shaded area shows the corresponding @xmath207 credible intervals .
the thinner dark gray line depicts the data histogram . ] . the solid line reports the estimated pointwise posterior means , the shaded area gives the corresponding @xmath207 credible intervals , while the dashed line shows the sample quantile differences .
data are cubic - root transformed . ]
we now describe the results for the gqte @xmath11 introduced in section [ estimation ] for some choices of the function @xmath4 .
we start from the qte , denoted as @xmath208 in , whose estimate is shown in figure [ delta_id ] .
the solid line represents the posterior mean of the medical costs qte between cases and controls and the gray area is the corresponding @xmath207 credible interval , while the dashed line portrays the sample quantile differences .
we can see that the distribution of the medical expenditures for subjects with smoking attributable diseases is always above that of those without smoking - related diseases . however , a much larger variability results in estimating the difference for the very extreme quantiles . this behavior is not too surprising since the two samples become very sparse as the medical expenditures become bigger ( see the boxplots in figure [ data_summary ] )
. figures [ delta_mu](a ) and ( b ) contain the posterior distributions of the ate and the standard deviation difference , as defined in section [ gqte ] .
the sample mean and standard deviation differences are depicted in the two plots with a vertical dotted line , while the @xmath207 credible intervals are indicated using dashed lines .
the ate estimated posterior mean ( on the log scale ) is equal to @xmath209 , while the estimated posterior mean of the standard deviation difference is @xmath210 .
these results prove that having a smoking attributable disease has a significant negative impact on both the location and scale of the single hospitalization medical cost distribution .
after re - transforming the estimated quantiles on the original scale , we get an estimated posterior mean for the ate between diseased and non - diseased subjects equal to @xmath211 . credible interval , while the dotted line is the sample mean difference .
( b ) estimated posterior distribution of the standard deviation difference between cases and controls medical expenditures ( cubic root transformed ) , as defined in section [ gqte ] .
the vertical dashed lines represent the @xmath207 credible interval , while the dotted line depicts the sample standard deviation difference . ] finally , figure [ delta_tw ] shows the impact of the treatment variable ( i.e. , having or not a smoking attributable disease ) on the tailweight functions @xmath212 , defined in , of the two populations .
the tails of the medical costs distribution for the diseased subjects tend to be heavier than those of the non - diseased ones for values of @xmath6 up to approximately @xmath213 , but the situation is inverted as we move to consider higher percentiles . hence , while the fact of being affected by smoking attributable diseases tends to increase both the average and the variance of the medical expenditures distribution , we have found that the opposite occurs to the tail probabilities , that is , to the chances of incurring very high medical costs in a single hospitalization . as a last comment , we would like to remark on the explicit choice we made to exclude the observations with null medical costs .
we took this decision because the inclusion of this further feature of the data requires the extension of our approach to a two - part modeling framework @xcite , which does nt appear to be straightforward in our context . between cases and controls , as defined in .
the solid line represents the estimated pointwise posterior means , while the shaded area shows the corresponding @xmath207 credible intervals . ]
in this paper we have introduced a new parameter , the gqte , for assessing the effect of a binary covariate on a response and a novel methodology to estimate it .
the gqte generalizes the most common approaches available in the literature , that is , the well - known average treatment effect ( ate ) and the quantile treatment effect ( qte ) , since it allows to evaluate the effect of a treatment on any arbitrary characteristic of the outcome s distributions under the two treatment conditions . to estimate the gqte we have proposed a bayesian procedure , where we assume that a monotone transformation of the quantile ratio is modeled as a smooth function of the percentiles
this assumption allows to increase efficiency by borrowing information across the two groups .
the idea of quantile ratio smoothing has first been introduced by @xcite . in the present work we extended that proposal in several ways : 1 ) we let the link between the quantile ratio and the percentiles be general and application - specific , allowing to take into account the tail heaviness of the distributions involved in the analysis ; 2 ) we derive a closed form expression for the model likelihood ; 3 ) our methodology is not limited to the mean difference between the treated and the controls , but provides a comprehensive assessment of the treatment effect ; 4 ) finally , we embed the whole estimation process within a bayesian framework allowing to make inference on the gqte @xmath48 for any choice of the function @xmath4 , and for both symmetric and highly skewed outcomes . the gqte is a marginal measure in the sense that it provides an estimate of the treatment effect over an entire population . in the econometrics literature
this kind of approach is usually termed the _ unconditional _
qte @xcite in contrast with the _ conditional _ qte , where the treatment effect is determined separately for different combinations of a set of covariates @xcite .
the inclusion of covariates can improve the efficiency of an estimator even when the primary goal of the analysis is a marginal effect .
accordingly , methods have been proposed to extract marginal quantiles from estimates of conditional quantiles @xcite .
a challenge in extending our approach along these lines is the lack of an `` iterated expectation '' result does imply @xmath214 , the same conclusion does nt hold for the conditional quantiles . ] for the quantiles ( see for example * ? ? ?
* chapter 7 ) . to further clarify our goals , we want to stress that in this paper no particular emphasis has been placed on the causality issues that naturally comes into play when the objective is the estimation of a treatment effect ( see for example * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
more precisely , our intent here is solely to provide a general measure of the effect of a binary treatment on a response variable , together with a flexible approach to estimate it .
we compared the performance of our estimation approach with other highly flexible methods in a simulation study for the mean difference between two populations .
our study revealed that the gqte performs generally better than the other competing estimators at least in estimating the mean difference .
we have applied our methodology to the nmes data set to assess the effect of being affected by smoking attributable diseases on the single hospitalization medical costs distribution .
we have found that having these diseases increases the average medical bill amount as well as its variability in the population , while it reduces the probability of incurring higher bills .
our approach can be extended in various directions .
the most promising research question we can see involves taking into account individual level characteristics in measuring the effect of a treatment . in our context
, this would involve the estimation of a conditional version of @xmath48 , something like @xmath215 . the clear advantage of including covariates would be an increase in the efficiency of the estimates @xcite . to control for systematic differences in covariates between two populations ,
a common strategy is to group units into subclasses based on covariate values , for example using propensity score matching , and then to apply our method within strata of propensity scores @xcite , as implemented for example in @xcite .
currently we are considering only a binary treatment effect , so another important line of research is the extension of the methods to categorical ordinal and to continuous treatments . a further direction for future research concerns the choice of the number of degrees of freedom @xmath30 .
in this paper we adopted the simple approach of choosing @xmath30 by minimizing an empirical version of the @xmath216 distance between the @xmath10 density estimate and its kernel density estimate ( see subsection [ dfselection ] ) .
more structured solutions can obviously be considered .
a natural extension would allow @xmath30 to be a random quantity to be estimated together with all the other parameters using a trans - dimensional mcmc approach , like for example the reversible jump algorithm @xcite .
while this solution would allow to take into account also the uncertainty connected to the a priori ignorance about the @xmath30 value , the consequence would be a dramatic increase in the computational workload of the estimation algorithm .
the research of dominici was supported by award number r01es012054 ( statistical methods for population health research on chemical mixtures ) from nih / niehs , award numbers r83622 ( statistical models for estimating the health impact of air quality regulations ) and rd83241701 ( estimation of the risks to human health of pm and pm components ) from epa , award number 4909-rfa11 - 1/12 - 3 ( causal inference methods for estimating long term health effects of air quality regulations ) from hei and award number k18 hs021991 ( a translational framework for methodological rigor to improve patient centered outcomes in end of life cancer research ) from ahrq .
the content is solely the responsibility of the authors and does not necessarily represent the official views of the above institutions .
together with the cases presented in section [ gqte ] , many other less conventional measures of the difference between two distributions can be obtained by properly choosing the @xmath4 function in the gqte definition .
for example , by choosing @xmath217 we obtain the difference between the population @xmath218-th moments @xmath219 using the fact that for a random variable @xmath220 with expected value @xmath221 , variance @xmath222 and quantile function @xmath223 it holds that ( see @xcite or @xcite ) @xmath224 ^ 2}\,dp=\int_0 ^ 1 { q(p)^{2 } } \,dp - \mu^{2 } , \ ] ] by suitably choosing the @xmath4 function , we recover the difference between the two population variances as @xmath225-\left[\int_0 ^ 1{q_2(p)^2}\,dp-\left ( \int_0 ^ 1{q_2(p)}\,dp\right)^2\right ] \nonumber\\ & = & \left[\int_0 ^ 1{q_1(p)^2}\,dp-\int_0 ^ 1{q_2(p)^2}\,dp\right ] -\left[\left(\int_0 ^ 1{q_1(p)}\,dp\right)^2-\left(\int _ 0 ^ 1{q_2(p)}\,dp\right)^2\right ] \nonumber\\ & = & \delta_{\mu^2}-\left(\mu_1 ^ 2-\mu_2 ^ 2\right),\end{aligned}\ ] ] however , the cases encompassed by the gqte include many other quantile - based indexes that are less frequently used in the literature , like the inter-@xmath6-range @xmath226 , or the skewness - ratio @xmath227/[q(0.5)-q(p)]$ ] , @xmath17 , which provide robust measures of the scale and shape of a distribution ( for a list of these indexes see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
a quantity of particular interest to economists is the difference between inter - decile ratios , defined as @xmath228 which is commonly used to measure the inequality in a population ( see * ? ? ?
the previous quantity can be easily generalized as follows @xmath229 for any @xmath230 .
notice that all these indexes are obtainable from the general definition by properly choosing the function @xmath4 . as a last example
, we consider a further gqte special case that is based on the so called _ tailweight function _ defined as @xmath231 which is used to quantify the probability allocated in the tails of a distribution .
one can compute the difference between the tailweight functions for two populations by choosing the logarithmic derivative of the quantile function as the @xmath4 functional in , that is @xmath232.\end{aligned}\ ] ] if @xmath233 , we can conclude that the treatment is causing a thickening of the @xmath47 distribution tails as compared to those of @xmath10 if @xmath233 . finally , note
that , thanks to the equivariance property of the quantiles , can be written also as @xmath234 \nonumber\\ & = & \frac{d}{dp } \delta_{\log}(p ) \,,\end{aligned}\ ] ] where @xmath235 , @xmath18 , indicates the quantile of the log - transformed data and @xmath236 denotes the parameter calculated on the quantiles of the log - transformed data .
[ proof_th1 ] differentiate with respect to @xmath6 to get @xmath237+x'(p,\lambda)\,{\boldsymbol}{\beta}\,q_2(p)\left\{\frac{d}{d\left ( x(p,\lambda)\,{\boldsymbol}{\beta}\right)}h^{-1}\left[x(p,\lambda)\,{\boldsymbol}{\beta}\right]\right\},\ ] ] where @xmath238 denotes the so called _ quantile density function _ for the population @xmath239 , while @xmath240 corresponds to the derivative of @xmath35 , @xmath17 ( properly resized because a constant is normally included in the design matrix @xmath35 ) .
apply now to both @xmath241 and @xmath242 the following relationship between the quantile density and density quantile functions ( see for example * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) @xmath243 to get the expression @xmath244 \\
+ x'(p,\lambda)\,{\boldsymbol}{\beta}\,q_2(p ) \left\{\frac{d}{d\left ( x(p,\lambda)\,{\boldsymbol}{\beta}\right)}h^{-1}\left[x(p,\lambda)\,{\boldsymbol}{\beta}\right]\right\},\end{gathered}\ ] ] and hence @xmath245}{1-f_1\left(q_1(p)|{\boldsymbol}{\eta}\right)\,x'(p,\lambda)\,{\boldsymbol}{\beta}\,q_2(p)\left\{\frac{d}{d\left(x(p,\lambda)\,{\boldsymbol}{\beta } \right)}h^{-1}\left[x(p,\lambda)\,{\boldsymbol}{\beta}\right]\right\ } } \,.\ ] ] finally , substituting in place of @xmath2 proves the main statement .
moreover , @xmath68 is a proper density function because : * @xmath246 , for any @xmath17 ; since @xmath247 and @xmath248 > 0 $ ] because , as assumed in , it is the ratio of two positive quantile functions , this fact can be proved by showing that the denominator of is nonnegative which is ensured by the constraint . *
@xmath249 , which is true because @xmath250 by construction .
a couple of immediate consequences of theorem [ theorem1 ] regard two cases that occur frequently in practice .
we provide the details about these situations in the next two corollaries .
let the same assumptions of theorem [ theorem1 ] hold .
suppose additionally that @xmath96 .
if for every @xmath17 the vector @xmath66 satisfies the constraint @xmath251 then the density quantile function @xmath252 for @xmath10 is @xmath253 let the same assumptions of theorem [ theorem1 ] hold .
suppose additionally that @xmath45 .
if for every @xmath17 the vector @xmath37 satisfies the constraint @xmath254 then the density quantile function @xmath252 for @xmath10 is given by @xmath255 note that in these two situations , the general constraint reduces to a linear constraint on @xmath142 .
_ case 1 : @xmath12 is uniform and @xmath94 .
_ here @xmath256 $ ] and @xmath96 . in this case
hence the density , distribution and quantile functions of @xmath12 are respectively @xmath257}\{y_1\ } \\
f_1(y_1|\theta_1 ) & = & \frac{y_1}{\theta_1 } \\
q_1(p|\theta_1 ) & = & \theta_1 p\ , , \qquad\qquad0<p<1\,.\end{aligned}\ ] ] from it follows that @xmath258}\{q_2(p)\,\beta_0\ } \ , \beta_0 \\
& = & \frac{\beta_0}{\theta_1}\,\mathbb{i}_{\left[0,\,\theta _ 1/\beta_0\right]}\{q_2(p)\}\,,\end{aligned}\ ] ] which is the density quantile function of a @xmath259 $ ] random variable with @xmath99 .
_ case 2 : @xmath12 is log - normal and @xmath106 $ ] . _
assume @xmath260 and @xmath45 .
in this case @xmath261 , where @xmath109 is the quantile function of a standard normal random variable . the density , distribution and quantile functions of @xmath12 are given by @xmath262 then by it follows @xmath263 ^ 2}{2}\right\ } } \ , \exp\left\{\,\mu_1+\sigma_1\phi ^{-1}(p)\right\ } \right ] } \\ & = & \frac{1}{\sqrt{2\pi}(\sigma_1-\beta_1)}\frac{\exp\left\ { -\frac{\left [ \phi^{-1}(p ) \right]^2}{2}\right\}}{\exp\left\{\ , ( \mu_1-\beta_0)+(\sigma_1-\beta_1)\phi^{-1}(p)\right\ } } \\ & = & \frac{1}{\sqrt{2\pi}(\sigma_1-\beta_1)}\,\frac{\exp\left\ { -\frac{\left [ ( \mu_1-\beta_0)+(\sigma_1-\beta_1)\phi ^{-1}(p)-(\mu_1-\beta_0 ) \right]^2}{2(\sigma_1-\beta_1)^2}\right\ } } { \exp\left\{\,(\mu_1-\beta_0)+(\sigma_1-\beta_1)\phi ^{-1}(p)\right\ } } \\ & = & \frac{1}{q_2(p)\sqrt{2\pi}(\sigma_1-\beta_1)}\,\exp\left\ { -\frac{\left [ \log q_2(p ) -(\mu_1-\beta_0 ) \right]^2}{2(\sigma _ 1-\beta_1)^2}\right\}\,,\end{aligned}\ ] ] which is the density quantile function of a @xmath111 random variable with @xmath112 and @xmath113 .
_ case 3 : @xmath12 is pareto and @xmath118 $ ] .
_ now @xmath119 and @xmath45 . in this case @xmath121
. the density , distribution and quantile functions of @xmath12 are given by @xmath264 then implies @xmath265^{-(a_{1}+1)}}{\exp\left\{-\beta_{0}-\beta_{1}\log(1-p)\right\ } \left\{1+a_{1}b_{1}^{a_{1}}\left[b_{1}(1-p)^{-\frac{1}{a_{1}}}\right ] ^{-(a_{1}+1)}\frac{\beta_{1}}{1-p}b_{1}(1-p)^{-\frac { 1}{a_{1}}}\right\ } } \\ & = & \frac{a_{1}}{a_{1}\beta_{1}+1}\left(b_{1 } e^{-\beta_{0}}\right ) ^{-1 } ( 1-p)^ { \frac{a_{1}\beta_{1}+1}{a_{1}}+1 } \\ & = & \frac{a_{1}}{a_{1}\beta_{1}+1}\left(b_{1 } e^{-\beta_{0}}\right ) ^{\frac{a_{1}}{a_{1}\beta_{1}+1 } } q_2(p)^{-\left(\frac { a_{1}}{a_{1}\beta_{1}+1}+1\right ) } \,,\end{aligned}\ ] ] which is the density quantile function of a @xmath266 random variable with @xmath124 and @xmath267 .
one can estimate the impact of a binary treatment on the @xmath218-th moments , denoted as @xmath268 in , by computing @xmath269 \right\}^r } \\ - \frac{1}{n_1}\sum^{n_1}_{i=1}{\left [ y_{1(i ) } \left\ { h^{-1}\left [ x\left(p_{1i},\lambda\right)\widehat{{\boldsymbol}{\beta}}^{(m ) } \right ] \right\}^{-1 } \right]^r } \,.\end{gathered}\ ] ] the last expression allows to estimate the treatment effect on the population variances , defined in , which is given by @xmath270 } \right)^2 \right .
. - \left ( \frac{1}{n_1}\sum^{n_1}_{i=1}{y_{1(i ) } \left\ { h^{-1}\left [ x\left(p_{1i},\lambda\right)\widehat{{\boldsymbol}{\beta } } ^{(m ) } \right ] \right\}^{-1 } } \right)^2 \right\ } \,.\end{gathered}\ ] ] as a concluding example , the effect of a binary treatment on the tailweight functions of two distributions , introduced in , can be obtained by first computing the posterior draws @xmath271 \right ) \right\ } , \ ] ] and then by applying . when @xmath45 , becomes @xmath272 and the estimate of @xmath273 is @xmath274 with @xmath275 , the posterior mean estimate of @xmath37 .
dominici , f. , zeger , s. l. , parmigiani , g. , katz , j. , and christian , p. ( 2006 ) .
`` estimating percentile - specific treatment effects in counterfactual models : a case - study of micronutrient supplementation , birth weight and infant mortality . ''
_ journal of the royal statistical society : series c ( applied statistics ) _ , 55 : 261280 . | we propose a new general approach for estimating the effect of a binary treatment on a continuous and potentially highly skewed response variable , the _ generalized quantile treatment effect _ ( gqte ) .
the gqte is defined as the difference between a function of the quantiles under the two treatment conditions . as such
, it represents a generalization over the standard approaches typically used for estimating a treatment effect ( i.e. , the average treatment effect and the quantile treatment effect ) because it allows the comparison of any arbitrary characteristic of the outcome s distribution under the two treatments . following @xcite
, we assume that a pre - specified transformation of the two quantiles is modeled as a smooth function of the percentiles .
this assumption allows us to link the two quantile functions and thus to borrow information from one distribution to the other .
the main theoretical contribution we provide is the analytical derivation of a closed form expression for the likelihood of the model . exploiting this result we propose a novel bayesian inferential methodology for the gqte .
we show some finite sample properties of our approach through a simulation study which confirms that in some cases it performs better than other nonparametric methods . as an illustration
we finally apply our methodology to the @xmath0 national medicare expenditure survey data to estimate the difference in the single hospitalization medical cost distributions between cases ( i.e. , subjects affected by smoking attributable diseases ) and controls .
./style / arxiv - ba.cfg , , |
with the rapid emergence of social networking websites , e.g. , facebook , twitter , linkedin , epinions , etc .
, a considerable amount of attention has been devoted to investigating the underlying social mechanisms in order to enhance users experiences @xcite@xcite@xcite@xcite .
traditional social network analysis concerns itself primarily with unsigned social networks such as facebook or myspace which can be modeled as graphs , with nodes representing entities , and positively weighted edges representing the existence of relationships between pairs of entities .
recently , signed directed social networks , in which the relationships between users can be either positive ( indicating relations such as trust ) or negative ( indicating relations such as distrust ) , are increasingly common . for instance , in epinions @xcite , which is a product review website with an active user community , users can indicate whether they trust or distrust other users based upon their reviews ; in slashdot @xcite@xcite , which is a technology - related news website , users can tag each other as `` friend '' or `` foe '' based upon their comments .
such a signed directed network can be modeled as a graph expressed as an asymmetric adjacency matrix in which an entry is @xmath0 ( or @xmath1 ) if the relationship is positive ( or negative ) and 0 if the relationship is absent .
one of the fundamental problems in signed social network analysis is edge sign inference @xcite@xcite , i.e. , inferring the unknown trust or distrust relationship given the existence of a particular edge . to address this issue ,
many approaches have been developed based upon two main social - psychological theories , i.e. , structural balance theory @xcite@xcite and social status theory @xcite .
structural balance theory is more well - known and it states that people in signed networks tend to follow the rules that `` the friend of my friend is my friend '' , `` the enemy of my friend is my enemy '' , etc .
social status theory , which is implicit in guha et al .
@xcite , further exploited by leskovec et al .
@xcite , and based upon a foundation in social psychology @xcite , considers a positive directed edge to indicate that the initiator of the edge views the recipient as having higher status and a negative directed edge to indicate that the recipient is viewed as having lower status .
the relative levels of status determine the allowed sign - direction pairs for an edge assuming that this edge exists .
although both structural balance theory and social status theory have proved useful for explaining the signs of edges in signed networks , neither is suitable for explaining an observed edge when the two nodes connected by this edge share no common neighbor ( e.g. , common friend ) , and in fact , structural balance theory simply does not apply to this situation .
since many real world social networking graphs tend to be very sparse , this is the case for a large fraction of their edges . to better explain the observed edge signs in general , in this paper we develop a novel approach to address this issue by applying a new model for node types . to summarize the contributions of this paper : * we explore the underlying local node structures in fully observed signed directed networks , recognizing that there are 16 different types of node and each type of node constrains both its incoming node types and its outgoing node types , i.e. , the signs of their edges must be consistent with their types . *
we show that node type features can be extended to incorporate structural balance theory or social status theory , to help make predictions for those edges whose endpoints have common neighbors .
* for the purpose of practical applications , we derive bayesian node features ( including bayesian node type and bayesian node properties ) based upon partially observed signed directed networks .
* we conduct empirical studies based upon three real world datasets and show that our proposed approach can outperform state - of - the - art algorithms .
in the past few years , many approaches have been developed to explore different aspects of signed networks , ranging from edge sign prediction @xcite@xcite@xcite@xcite to community detection @xcite@xcite .
most of these approaches are based upon structural balance theory or social status theory .
the investigation of signed networks @xcite@xcite@xcite@xcite can be traced back to the 1920s .
heider @xcite first formulated structural balance theory within social psychology .
after that , cartwright and harary @xcite formally provided the notion of structural balance with undirected triads ( as shown in figure [ fig : fig2 ] ) and proved its necessity and sufficiency by utilizing the mathematical theory of graphs .
intuitively , their theory can be explained as : `` the friend of my friend is my friend '' ( @xmath2 ) , `` the enemy of my friend is my enemy '' ( @xmath3 ) , `` the friend of my enemy is my enemy '' ( @xmath3 ) , and `` the enemy of my enemy is my friend '' ( @xmath3 ) .
conceptually , their theory claims that @xmath2 and @xmath3 are balanced while @xmath4 and @xmath5 are unbalanced .
davis @xcite further generalized this theory to weak structural balance theory by allowing all the edges of triads to be negative , i.e. , `` the enemy of my enemy is my enemy '' ( @xmath5 is also balanced ) .
note that these two balance theories were initially intended for modeling undirected networks , although they have been commonly applied to directed networks by disregarding the direction of edges @xcite .
guha et al .
@xcite first considered the edge sign prediction problem by developing a trust propagation framework to predict the trust ( or distrust ) between pairs of nodes . in their framework , they calculate a combined matrix which is a linear combination of four different one - step propagations , i.e. , direct propagation , co - citation , transpose trust , and trust coupling .
then the trust and distrust propagations are achieved by calculating a linear combination of powers of this combined matrix .
a shortcoming of this approach is that it can not be explained by structural balance theory @xcite@xcite . motivated by this trust propagation idea @xcite and informed by social psychology @xcite , leskovec et al .
@xcite developed social status theory to explain signed directed networks .
in this theory , they assume that if there is a positive edge from @xmath6 to @xmath7 , it represents the fact that @xmath6 regards @xmath7 as having higher status than himself ( or herself ) , and if there is a negative edge from @xmath6 to @xmath7 , it represents the fact that @xmath6 regards @xmath7 as having lower status than himself ( or herself ) . assuming everyone in the system agrees on the same status ordering , we can infer signs easily as long as the existence and direction of edges are available .
when prior status information for @xmath6 and @xmath7 is not available , we can still perform sign inference using the context provided by the rest of the network .
for instance , in figure [ fig : fig3 ] , the sign of @xmath6 to @xmath7 can be inferred by referring to the status of @xmath8 , and is unambiguous in half the cases . based upon structural balance theory and social status theory , leskovec et al .
@xcite selected degree features and directed triad features for edges in signed directed networks .
specifically , for the edge from node @xmath6 to node @xmath7 , they consider seven degree features , i.e. , @xmath9 and @xmath10 , the number of incoming positive and negative edges to @xmath7 , respectively ; @xmath11 and @xmath12 , the number of outgoing positive and negative edges from @xmath6 , respectively ; @xmath13 , the number of common neighbors ( i.e. , embeddedness ) of node @xmath6 and node @xmath7 ; @xmath14 and @xmath15 , the total out - degree of @xmath6 and the total in - degree of @xmath7 , respectively .
since each of the 16 triad types in figure [ fig : fig3 ] provides different evidence for the sign of the edge from node @xmath6 to node @xmath7 , directed triad features of this edge are encoded in a 16-dimensional vector counting the number of triads of each type in which this edge is involved . after computing the degree or directed triad features for the edge from @xmath6 to @xmath7 , a logistic regression classifier is used to combine the evidence from these individual features into an edge sign prediction . subsequently , chiang et al .
@xcite extended this approach by considering longer cycles ( e.g. , quadrilaterals , pentagons ) while ignoring the directions of edges to reduce the computational complexity .
hsieh et al .
@xcite formulated the sign inference problem as a low rank matrix completion ( approximation ) problem based upon weak balance theory .
note that this approach was originally developed to explain a signed undirected network which is associated with a symmetric adjacency matrix and is different from our setting ( signed directed networks ) in this paper .
we remark that structural balance theory is also popular for community detection in signed networks @xcite@xcite .
although many approaches based upon structural balance theory or social status theory have been developed to perform edge sign prediction in signed networks , they can not work well when few topological features , i.e. , undirected ( or directed ) triads and long - range cycles , are available in the network .
since many real world signed directed social networking graphs are very sparse , the efficacy of methods based upon these theories is limited .
a more general approach for such networks is necessary .
in this paper , we consider three well - known signed directed social networks : wikipedia @xcite , slashdot @xcite@xcite and epinions @xcite : * the wikipedia data comprise a voting network for promoting candidates to the role of admin .
the voters , half coming from existing admins and another half coming from ordinary wikipedia users , can indicate a positive ( for supporting ) or negative ( for opposing ) vote with respect to the promotion of a candidate @xcite .
* slashdot is a social website focusing on technology related news . in slashdot zoo , users can tag each other as friends ( like ) or foes ( dislike ) based upon comments on articles .
* epinions , which is a product review website , is a trust network in which users can indicate whether they trust or distrust each other based upon their reviews .
the detailed statistics of these datasets are provided in table [ tab : table1 ] .
note that in all three datasets , the majority of the edges is positive . due to this imbalance ,
simply predicting all edges to be positive would yield 78.78@xmath16 , 77.4@xmath16 , and 85.0@xmath16 accuracy across the three datasets . to show the effectiveness of any approach
, it should achieve substantially better performance than this .
[ fig : subfig4a ] [ fig : subfig4b ] [ fig : subfig4c ] figure [ fig:4 ] shows the fraction of edges versus edge embeddedness @xcite ( the number of common neighbors of the two nodes connected by the edge ) for three datasets .
we observe that the edges with zero embeddedness comprise about @xmath17 , @xmath18 , @xmath19 of the edges for wikipedia , slashdot , and epinions , respectively .
note that a large fraction of zero embeddedness edges means that triad features @xcite can not work well for edge sign prediction .
this is because the entries of triad feature vector will be zero and thus the triad features provide no evidence for edge sign prediction .
in this paper , a fully observed signed directed network refers to a network in which there is no uncertainty about the existence of any directed edge and its associated sign .
we consider a fully observed signed directed network as a graph @xmath20 , where @xmath21 is the vertex set of size @xmath22 , @xmath23 is the edge set of size @xmath24 , and @xmath25 is the associated signed adjacency matrix . because @xmath26 is a directed network , @xmath27 is an asymmetric matrix and can be represented as : @xmath28 note that @xmath29 represents no directed edge from node @xmath30 to node @xmath31 . in this section ,
we first investigate local node structures within fully observed signed directed networks , recognize a set of node types , show that these node types can be used to explain real world signed directed social networks , and show how to encode a specific node in such networks .
next , we explore how these node types interact with one another and how these interactions can explain the edge signs .
finally , we show our approach can be extended to incorporate structural balance theory or social status theory . in our study
, we focus on analyzing signed directed networks because they are more general and common than signed undirected networks in real world applications .
for instance , each of the three datasets we consider in this paper is a signed directed network .
generally , signed directed networks are sparse graphs in which nodes may be categorized into 4 groups based upon whether they have incoming edges and outgoing edges , i.e. , nodes with neither incoming nor outgoing edges ( e.g. , @xmath32 in figure [ fig : fig5 ] ) , nodes with only incoming edges ( e.g. , @xmath33 , @xmath34 , and @xmath35 in figure [ fig : fig5 ] ) , nodes with only outgoing edges ( e.g. , @xmath36 , @xmath37 , and @xmath38 in figure [ fig : fig5 ] ) , and nodes having both incoming and outgoing edges ( e.g. , @xmath39 , @xmath40 , @xmath41 , etc . ) . moreover ,
both the incoming edges and the outgoing edges of a given node can be categorized into 3 classes , one class with only positive edges , another class with only negative edges , and the third class with a mixture of positive and negative edges .
combining these two principles , the nodes in signed directed networks can be categorized into 16 types , shown in figure [ fig : fig5 ] . note that the edges in figure [ fig : fig5 ] only indicate the types of the incoming or outgoing , i.e. , they do not represent the actual nonzero number of incoming ( outgoing ) positive ( negative ) edges .
the fractions of each node type for the three real world datasets are shown in figure [ fig : fig6 ] . to represent each node effectively , in addition to its node type @xmath42
, we should consider its associated node properties , i.e. , the relative level of the number of positive ( negative ) incoming edges @xmath43 @xmath44 , and the relative level of the number of positive ( negative ) outgoing edges @xmath45 ( @xmath46 ) .
first , we can use a 16-dimensional binary vector , @xmath47,@xmath48 , , @xmath49 $ ] , to indicate the node type .
the indicator @xmath50 is 1 if @xmath42 is the same as @xmath51 .
next , we can use a vector @xmath52 @xmath53 $ ] to denotes the ratio of positive ( negative ) incoming edges and that of positive ( negative ) outgoing edges .
intuitively , @xmath54 , @xmath55 , @xmath56 , and @xmath57 represent the locally propagating properties of a node ( node property ) and they can be calculated with @xmath58 @xmath59 @xmath60 @xmath61 where we set @xmath62 to avoid zero denominators .
figure [ fig : fig7 ] shows examples of these two parts of features for node type @xmath63 , @xmath64 , and @xmath41 .
notice that @xmath65 if there is any input edge , but this sum is zero if there are none ; so these features are not redundant . also note that node property is essentially different from degree features because degree features aim to model a particular edge by considering the initiator s outgoing edges , recipient s incoming edges , and their common neighbors .
note that although the node properties , i.e. , @xmath66@xmath67 $ ] implicitly indicate the node type information , it is still useful to consider type indication , i.e. , @xmath47,@xmath48 , , @xmath49 $ ] .
since the latter is not a linear combination of the former , it can provide non - redundant information in the logistic regression classifier we will describe in the section 6 .
we have shown there are 16 possible node types in any signed directed network .
hence , theoretically there are @xmath68 combinations of node types . given a node of a certain type , however
, it usually can only connect to ( or be reached from ) nodes in a subset of these types due to the compatibility of both directions and signs .
in other words , there exists a logic to determine whether two nodes can be reached or not and whether the sign should be positive or negative .
for instance , given a node of type @xmath69 , it can only be connected from a node of type @xmath70 or @xmath41 and the edge sign can only be negative .
similarly , given a node of type @xmath71 , it can only be connected from a node of type @xmath72 or @xmath41 and connected to a node of type @xmath73 , or @xmath41 . moreover ,
the edge sign is determined as positive and negative , respectively .
given an edge from @xmath6 to @xmath7 , based upon the combinations of node type @xmath6 and node type @xmath7 , this edge can be categorized into three classes .
@xmath74 denotes the edge sign is determined to be positive , @xmath75 denotes the edge sign is determined to be negative , and @xmath76 denotes the edge sign that can not be determined by the interaction of current two node types , i.e. , the edge sign can be either positive or negative . in our three datasets , there are @xmath77 , @xmath78 and @xmath79 determined edges ( i.e. , @xmath74 and @xmath75 ) for wikipedia , slashdot and epinions , respectively , each a large fraction of the total edges .
although node types have shown their effectiveness for explaining the edge signs in fully observed signed social networks , there exists a fraction of the total edges for which signs can not be explained simply by node types . in this case
, we can incorporate structural balance or social status theory with node types to address this issue . as we described the node types and the interactions of these node types in the previous subsections
, we did not need to consider whether there is any common neighbor for a pair of nodes .
we should , however , be aware that when common neighbors exist for a pair of nodes , structural balance or social status theory may help to explain the sign of an edge between them . in figure
[ fig : fig8 ] , for instance , since @xmath80 has both positive and negative outgoing edges and @xmath41 has both positive and negative incoming edges , the sign of the edge between them can not be determined by the interaction of these two types .
since @xmath80 and @xmath41 have two common neighbors , however , the sign of the edge between them may be explained by either structural balance theory or social status theory .
within structural balance theory , we can disregard the directions of these two triads . from the red ( dotted ) triad
, we can infer the sign of the edge between @xmath80 and @xmath41 to be positive based upon the rule that `` my friend s friend is my friend '' . from the blue ( dashed ) triad , on the contrary
, we can infer the sign of this edge to be negative based upon the rule that `` my friend s enemy is my enemy '' .
within social status theory , since both the red ( dotted ) triad and the blue ( dashed ) triad indicate that @xmath80 has higher status than @xmath41 , they consistently imply that the sign of the edge from @xmath80 to @xmath41 is negative . to incorporate
these directed triads as features , we use the same approach as leskovec et al .
@xcite@xcite . given an edge from @xmath6 to @xmath7 , and a common neighbor @xmath8 of @xmath6 and @xmath7
, the edge between @xmath6 and @xmath8 can have four possible configurations , i.e. , @xmath81 , @xmath82 , @xmath83 , and @xmath84 .
similarly , there are four possible signed edges between @xmath8 and @xmath7 .
hence we can obtain 16 types of triads each of which may provide different evidence about the sign of the edge from @xmath6 to @xmath7 .
in real world applications , signed directed networks are often partially observed , i.e. , several edges signs are unknown or hidden . for example , in the wikipedia dataset , we probably know that someone has voted on a candidate , but we may not know this voter s opinion . in this case , we would like to infer this voter s opinion by learning some patterns based upon observed edges in the network .
however , when these unobserved edges take different signs , both the node types and node properties may change . in this case , simple node features ( including node types and node properties ) may not be capable of capturing the range of possible unobserved signs and thus will be not reliable . to address this issue ,
we extend node features to bayesian node features by considering prior knowledge about unobserved signs in partially observed signed directed networks . similarly to a
fully observed signed directed network , a partially observed signed directed network can also be represented as a graph @xmath20 , where @xmath21 is the vertex set of size @xmath22 , @xmath23 is the edge set of size @xmath24 , and @xmath25 is the associated signed adjacency matrix . since @xmath26 is a directed network , @xmath27 is an asymmetric matrix and can be represented as : @xmath85 @xmath29 represents no directed edge from node @xmath30 to node @xmath31 . in this section ,
we first introduce bayesian node type , show how to calculate it based upon both observed ( training ) edges and unobserved ( test ) edges , and present two ways to encode the interaction of bayesian node types .
next , we show how to calculate and represent bayesian node properties . given any node in a partially observed network
, assuming @xmath86 denotes the number of unobserved incoming edges and @xmath87 denotes the number of unobserved outgoing edges , let @xmath88 ( or @xmath89 ) represent the prior probability of incoming edges being positive ( negative ) and @xmath90 ( or @xmath91 ) represent the prior probability of outgoing edges being positive ( negative ) .
from these probabilities , we can calculate its probability distribution over node types .
take a node of type @xmath36 for example ( as shown in figure [ fig : fig9 ] ) , if @xmath92 and @xmath93 , its node type does not change ; if @xmath92 and @xmath94 , it has @xmath95 probability to be @xmath36 and @xmath96 probability to be @xmath38 ; if @xmath97 and @xmath93 , it has @xmath98 probability to be @xmath99 , @xmath100 probability to be @xmath101 , and @xmath102 probability to be @xmath103 ; if @xmath97 and @xmath94 , it has @xmath104 probability to be @xmath99 , @xmath105 probability to be @xmath101 , @xmath106 probability to be @xmath103 , @xmath107 probability to be @xmath80 , @xmath108 probability to be @xmath40 , and @xmath109 @xmath110 probability to be @xmath41 .
this calculation determines a 16 dimensional vector which encodes the distribution of possible node types .
note that similar vectors can be calculated for other types of nodes .
we do not specify the calculation for each node type due to the the space limit . to initialize @xmath88 and @xmath90 , we can also use the bayesian node properties @xmath111 and @xmath112 , i.e. , @xmath113 and @xmath114 , to claim that each unobserved edge obeys bayesian node properties ( i.e. , local priors ) .
given an observed ( training ) edge connecting node @xmath6 and node @xmath7 , we can obtain two vectors @xmath115 and @xmath116 by calculating their bayesian node types . to encode the interaction of bayesian node types
, we can ( 1 ) simply concatenate these two vectors to form a 32 dimensional vector ; or ( 2 ) calculate the kronecker product of these two vectors , i.e. , @xmath117 , and form a 256 dimensional vector
. we should be aware that the vector formed by the kronecker product encodes the probability distribution of different node type interactions .
given an unobserved ( test ) edge connecting node @xmath6 and node @xmath7 , we should consider both possible signs as shown in figure [ fig : fig10 ] .
specifically , we first decompose this kind of interaction into two separate cases , i.e. , the edge sign being positive and negative .
next , we calculate the bayesian node types , represent their interactions ( either concatenation vector or kronecker product vector ) of both cases . finally , we calculate the linear combination of these two cases with respect to the prior probability of the signs ( i.e. , @xmath118 and @xmath119 ) over observed edges . given a node in a partially observed network , assuming @xmath86 denotes the number of unobserved incoming edges and @xmath87 denotes the number of unobserved outgoing edges , by assigning to these @xmath120 edges different signs , the node properties also change . to capture the range of possible signs of unobserved edges
, we should consider bayesian node properties , i.e. , incorporating prior information , namely the expected number of incoming positive ( negative ) and outgoing positive ( negative ) edges , with the number of positive ( negative ) incoming edges @xmath43 @xmath121 or outgoing edges @xmath122 @xmath123 .
specifically , the bayesian node properties are represented as following : @xmath124 @xmath125 @xmath126 @xmath127 where @xmath118 is the prior probability of positive edges and @xmath119 is the prior probability of negative edges . to encode the interaction of bayesian node properties , we simply concatenate two of the bayesian node properties vectors to form an 8 dimensional vector . as in the previous section ,
when common neighbors exist for a pair of nodes , structural balance or social status theory may be incorporated with bayesian node features to help explain the sign of an edge between them .
given a fully observed signed directed network , the node type interactions , extended to triads with structural balance or social status theory , are useful to explain the edge signs . partially observed signed directed networks , however , are too complicated to fully conform to the rules of simple node type interactions .
also , as illustrated in figure [ fig : fig8 ] , if there are multiple common neighbors for @xmath80 and @xmath41 , structural balance theory ( or social status theory ) may conflict with itself . to address this issue
, we can utilize a logistic regression to combine the evidence from the interaction of bayesian node features and triad features .
we now consider the features collected for the logistic regression .
the features we utilize can be divided into three classes .
one class comes from bayesian node type interaction ( 32 or 256 dimensional vector ) ; another class is based upon bayesian node properties interaction ( 8 dimensional vector ) ; the last class is triads ( 16 dimensional vector ) . given a partially observed signed directed social network , we first use a logistic regression to fit the features of observed edges ( training data ) and then utilize the learned coefficients to linearly combine the evidence from each individual feature of unobserved edges ( test data ) so as to predict the sign .
the logistic regression can be written in the following form @xmath128}\end{aligned}\ ] ] where @xmath129 is the label , @xmath0 represents positive edge while @xmath130 represents negative edge .
@xmath131 is the feature vector , and @xmath132\in\mathbb{r}^{d+1}$ ] are the coefficients we estimate from the features of observed edges ( training data ) .
[ fig : subfig10a ] [ fig : subfig10b ] [ fig : subfig10c ]
in this section , we conduct empirical studies based upon wikipedia @xcite , slashdot @xcite@xcite and epinions @xcite . we first construct three fully observed asymmetric adjacency matrices ( as in eq.(1 ) ) based upon these three datasets .
next , for each adjacency matrix , we randomly remove 10@xmath16 of edges signs and form a partially observed network ( as in eq .
subsequently , we calculate bayesian node features ( including bayesian node types and bayesian node properties ) and triad features for both observed ( training ) and unobserved ( test ) edges . then , we estimate the parameters of logistic regression based upon the features of observed edges and make predictions based upon the features of unobserved edges . in our experiment , we repeat this procedure 5 times and report the average prediction accuracy and standard deviation for each approach .
the baseline approaches are implemented with identical parameter settings as in the original works for fair comparisons .
.step by step justification results on wikipedia , slashdot , and epinions ( accuracy : @xmath16 and standard deviation ( std ) : @xmath16 ) .
[ cols="<,^,^,^",options="header " , ] in table [ tab : table7 ] , we compare bntc+bnp+triad and bntk+bnp+triad with the other five state - of - the - art approaches . we observe that bntc+bnp+triad and bntk+bnp+triad consistently outperform the other five algorithms .
note that these two variants achieve best accuracies of @xmath133 , @xmath134 , and @xmath135 over wikipedia , slashdot and epinions , respectively .
this is because these two variants not only can explain the edge signs well when common neighbors are not available but also can effectively explain the edge signs when common neighbors exist . in figure
[ fig : fig11 ] , we compare bntk+bnp+triad with degree features , triad features , and degree@xmath136triad features at different levels of embeddedness ( the number of common neighbors ) . in general
, we observe that when the minimum embeddedness increases , the performance of bntk+bnp+triad increases .
this is because structural balance theory and social status theory are incorporated into bntk+bnp+triad in the form of triads ( triad ) and are effective in explaining edge signs when common neighbors exist .
moreover , we notice that bntk+bnp+triad generally outperforms other methods with different levels of embeddedness .
this is because bntk+bnp+triad leverages the power of node type interactions as well as the power of structural balance or social status theory in the form of triad features ( triad ) .
we conduct cross - dataset evaluation with degree features , triad features , degree@xmath136triad features , and bayesian node features plus triad features in the form of bntk@xmath136bnp@xmath136triad on these three datasets .
the aim is to examine the generalization capability of each approach . in particular , given each type of features , we train them on one dataset ( e.g. , wikipedia ) and evaluate the edge sign prediction performance on another ( e.g. , slashdot ) . for each pair of datasets ,
the test is conducted 5 times based upon the random selected test sets .
we report the average accuracies of different approaches in table [ tab : table8 ] .
we observe that bayesian node features plus triad features in the form of bntk@xmath136bnp@xmath136triad can achieve the best performance on each pair of the cross - dataset evaluation .
this illustrates that bayesian node features plus triad features not only are useful on intra - dataset evaluation but also have good generalization capability .
this is extremely helpful for edge sign prediction in signed networks with few training examples .
in this paper , we explored the underlying local node structures in signed networks , recognizing that there are 16 different types of node and each type of node constrains both its incoming node types and its outgoing node types , i.e. , the sign of an edge between two nodes must be consistent with their types . this is a highly structured alternative to the ordered scalar node types postulated by social status theory .
we demonstrated that the interaction between these more complicated node types can explain edge signs well .
we also showed that our approach can be extended to incorporate triad features whose inclusion is motivated by structural balance theory or social status theory .
we derived bayesian node features ( including bayesian node type and bayesian node properties ) based upon partially observed signed directed network .
empirical studies based upon three large scale datasets , i.e. , wikipedia , slashdot , and epinions showed that the proposed bayesian node features plus triad features outperform state - of - the - art algorithms on edge sign prediction .
moreover , we showed that bayesian node features plus triad features are more effective than baseline approaches for cross - dataset edge sign predictions . in the future
, it will be interesting to study the link recommendation problem based upon bayesian node features as well as other explicit topological features of signed social networks .
this work was partially supported by the minerva research initiative under aro grants w911nf-09 - 1 - 0081 and w911nf-12 - 1 - 0389 .
j. jiang , c. wilson , x. wang , p. huang , y. dai , and b. y. zhao , _ understanding latent interactions in online social networks _ ,
proceedings of the acm the internet measurement conference , november , 2010 .
j. kunegis , s. schmidt , a. lommatzsch , j. lerner , e. w. deluca , and s. albayrak , _ spectral analysis of signed graphs for clustering , prediction and visualization _ , proceedings of the siam sdm , pp.559 - 570 , 2010 . | signed directed social networks , in which the relationships between users can be either positive ( indicating relations such as trust ) or negative ( indicating relations such as distrust ) , are increasingly common . thus the interplay between positive and negative relationships in such networks has become an important research topic .
most recent investigations focus upon edge sign inference using structural balance theory or social status theory .
neither of these two theories , however , can explain an observed edge sign well when the two nodes connected by this edge do not share a common neighbor ( e.g. , common friend ) . in this paper
we develop a novel approach to handle this situation by applying a new model for node types .
initially , we analyze the local node structure in a fully observed signed directed network , inferring underlying node types
. the sign of an edge between two nodes must be consistent with their types ; this explains edge signs well even when there are no common neighbors .
we show , moreover , that our approach can be extended to incorporate directed triads , when they exist , just as in models based upon structural balance or social status theory .
we compute bayesian node types within empirical studies based upon partially observed wikipedia , slashdot , and epinions networks in which the largest network ( epinions ) has 119k nodes and 841k edges .
our approach yields better performance than state - of - the - art approaches for these three signed directed networks . signed directed social networks ; node types ; bayesian node features ; edge sign prediction . |
synchronization of nonlinear oscillators is widely studied in physical and biological systems @xcite for underlying interests ranging from novel communications strategies @xcite to understanding how large and small neural assemblies efficiently and sensitively achieve desired functional goals @xcite .
the analysis of biological systems may , beyond their intrinsic interest , often provide physicists with novel dynamical systems possessing interesting properties in their component oscillators or in the nature of the interconnections .
we have presented our analysis of the experimental synchronization of two biological neurons as the electrical coupling between them is changed in sign and magnitude @xcite .
subsequent to that analysis we have developed computer simulations of the dynamics of the neurons which are based on conductance based hodgkin - huxley ( hh ) @xcite neuron models .
these numerical simulations quantitatively reproduced the observations in the laboratory @xcite .
the study of isolated neurons from the stomatogastric ganglion ( stg ) of the california spiny lobster _ panulirus interruptus _ using tools of nonlinear time series analysis @xcite shows that the number of active degrees of freedom in their membrane potential oscillations typically ranges from three to five @xcite .
the appearance of low dimensional dynamics in this biological system led us to develop models of its action potential activity which are substantially simpler than the hh models we and others @xcite have used to describe these systems .
we adopted the framework established by hindmarsh and rose ( hr ) @xcite in which the complicated current voltage relationships of the conductance based models are replaced by polynomials in the dynamical variables , and the coefficients in the polynomials are estimated by analyzing the observed current / voltage curves for the neurons under study .
building on biological experiments and on numerical analysis of models for the oscillations of isolated neurons , we have constructed low dimensional analog electronic neurons ( ens ) whose properties are designed to emulate the membrane voltage characteristics of the individual neurons . using these simple
, low dimensional ens we report here their synchronization and regularization properties , first when they are coupled electrically as the sign and magnitude of the coupling is varied , and then when they are coupled by excitatory and inhibitory chemical synapses .
we have also studied the behavior of an hybrid system , i.e. , one en and one biological neuron coupled electrically . as our models were developed on data acquired from biological neurons in synaptic isolation , the results we present here on pairs of interacting ens and hybrid systems serve to provide further confirmation of the properties of those model neurons , numerical and analog .
we have studied and built three dimensional and four dimensional models of hr type having the form @xmath0 where @xmath1 and @xmath2 are the constants which embody the underlying current and conductance based dynamics in this polynomial representation of the neural dynamics .
@xmath3 is the membrane voltage in the model , @xmath4 represents a `` fast '' current in the ion dynamics , and we choose @xmath5 , so @xmath6 is a `` slow '' current . taken alone the first three equations of the model can reproduce several modes of spiking - bursting activity observed in stg cells .
the first three equations were used in analog realization for our earlier experiments with 3d ens @xcite the equation for @xmath7 represents an even slower dynamical process ( @xmath8 ) , and it is included because a slow process such as the calcium exchange between intracellular stores and the cytoplasm was found to be required in hodgkin - huxley modeling @xcite to fully reproduce the observed chaotic oscillations of stg neurons .
both the three dimensional and four dimensional models have regions of chaotic behavior , but the four dimensional model has much larger regions in parameter space where chaos occurs , presumably for many of the same reasons the calcium dynamics gives rise to chaos in hh modeling .
the calcium dynamics is an additional degree of freedom with a time constant three times slower than the characteristic bursting times . in our analog circuit realization of the en model we used @xmath9 and @xmath10 .
the implementation of these constants in analog circuits always has about 5% tolerance in the components .
the main parameters we used in controlling the modes of spiking and bursting activity of the model are the dc external current @xmath11 and the time constants @xmath12 and @xmath13 of the slow variables .
figure [ 4dphases ] shows a chaotic time series of the four variables using the parameters above .
note how @xmath14 modulates the length of the bursts in @xmath15 .
each local minimum in the global oscillations of @xmath14 coincides with a short burst period .
the complexity achieved by the addition of @xmath14 can be observed in the projections of @xmath16 space to various three - dimensional spaces , @xmath17 and @xmath18 respectively , as shown in figure [ 4dphases ] .
table 1 presents the lyapunov exponents @xmath19 calculated from the vector field @xcite of equation ( [ eq : hrmodel ] ) for the 3d and 4d ens .
a positive lyapunov exponent is present in both models , indicating conclusively that they are oscillating chaotically . from this spectrum of lyapunov exponents
, we can evaluate the lyapunov dimension @xmath20 which is an estimate of the fractal dimension of the strange attractor for the ens @xcite .
the lyapunov dimension is defined by finding that number of lyapunov exponents @xmath19 satisfying @xmath21 then @xmath20 is defined as @xmath22 @xmath20 for each en is displayed in the last column table 1 . [ cols="^,^,^,^,^,^",options="header " , ] table 1 : lyapunov exponents @xmath19 and lyapunov dimension @xmath20 calculated from the vector field ( equation ( [ eq : hrmodel ] ) ) for the 3d and 4d electronic neuron models . as a reminder to the reader : the sum of all lyapunov exponents must be negative , and this is so for our results . also , one lyapunov exponent must be 0 as we are dealing with a differential equation .
we designed and built an analog electronic circuit which integrates equation ( [ eq : hrmodel ] ) .
we chose to build an analog device instead of using numerical integration of the mathematical model on the cpu of a pc or on a dsp board because digital integration of these equations is a slow procedure associated with the three different time scales in the model .
furthermore , a digital version of an en requires digital to analog and analog to digital converters to connect the model to biological cells .
analog circuits are small , simple and inexpensive devices ; it is easy to connect them to a biological cell , as we discuss below ( see also @xcite ) . in a practical sense
nearly an unlimited number of them can work together in real - time experiments .
finally , looking ahead to the construction of real - time networks of large number of these neurons , analog implementation is a necessity .
the block diagram of the analog circuit we use to represent the three dimensional and the four dimensional ens is shown in figure ( [ 4dcircuit ] ) .
it includes four integrators indicated by @xmath23 , two multipliers , two adders , and two inverters .
we used off - the - shelf general purpose operational amplifiers , national instruments model tl082 , to build the integrators , adder and inverter and used analog devices model ad633 as analog multipliers .
the integrator at the top of the diagram receives all components of @xmath24 , e.g. @xmath25 , @xmath26 , etc .
it has an additional input ( called @xmath27 ) which can be used for connections with other circuits or neurons .
the integrators invert the sign of their input , so the output signal will be @xmath28 multiplied by a time constant @xmath29 chosen to make the en oscillate on the same time scale as the biological neurons . the signal @xmath28 is used to generate the nonlinear functions @xmath30 and @xmath31 and these values go to the second and third integrators .
similarly , the other integrators generate voltages proportional to @xmath4 , @xmath32 , and @xmath7 .
a renormalization was made in the rest of the time constants in the circuit to make @xmath33 .
note that this rescaling is responsible for the different amplitudes in the numerical ( figure [ 4dphases ] ) and analog ( figures [ free ] , [ posel ] , [ negel ] , [ exchem ] , [ inchem ] ) experiments .
this circuit design allows us to easily switch from a three dimensional to a four dimensional model of the neuron .
we can connect or disconnect one wire , indicated as point a in figure [ 4dcircuit ] , to enable or disable the circuit block shown in the rectangle with a dashed outline . in equation([eq
: hrmodel ] ) this corresponds to setting @xmath34 in the @xmath35 equation .
the block indicated as na in figure [ 4dcircuit ] is an adjustable nonlinear amplifier .
we use it to rescale and change the shape of the output signal @xmath3 .
it can shrink or stretch different parts of the waveform , change the amplitude and move the trace as a whole up or down .
this shape adjustment is particularly important in experiments with groups of biological and electronic neurons interconnected with each other .
living neurons , even taken from the same biological structure , may generate considerably different waveforms .
the relative size of spikes and the interburst hyperpolarization is variable from cell to cell . in our circuits
we can precisely adjust the waveform of the en to be very close to that of each biological neuron in our experiments .
another reason to use circuits with variable waveforms is that it opens up the possibility of studying how the action potential waveforms affect the interactions among the neurons , electronic and biological .
indeed , the ability to vary the details of the waveforms provides an interesting handle on design of biometric circuitry for a variety of applications .
in living nervous systems one finds three general types of synaptic connections among neurons @xcite : ohmic electrical connections ( also called gap junctions ) and two types of chemical connections , excitatory and inhibitory . for our studies of the interconnections among ens and among ens and biological neurons @xcite , we built electronic circuits to emulate excitatory and inhibitory synaptic connections as well as the ohmic electrical connections .
the stg neural circuits are dominated by inhibitory interconnections and by ohmic electrical connections .
we now describe how we implemented each , and then we turn to the results of our synchronization experiments with these network connections .
we implemented an electrical synapse @xcite between the ens by injecting into one of the neurons ( @xmath36 ) a current proportional to the voltage difference between the two membrane potentials of the ens and into the other neuron ( @xmath37 ) injecting the same current but with the opposite strength .
the current into @xmath36 is @xmath38 + while @xmath39 we chose the dimensionless synaptic strength @xmath40 in the range @xmath41 .
$ ] over this range we observed the effects of positive and negative electrical coupling on the spiking and bursting behavior of the ens .
we recorded the electrical voltage signals corresponding to the membrane potentials of the ens using an analog to digital converter with a sampling rate of 5 khz . for each value of @xmath40 we waited at least 40 seconds to avoid transient dynamics and then recorded a data series 20 seconds long .
natural biological networks do not have negative conductance electrical coupling .
using an active device placed between the neurons we implemented negative electrical couplings in our experiments on two electrically coupled biological neurons as reported in @xcite .
to compare the results of our work there with the properties of coupled ens , we use negative coupling here as well .
we first implemented mutual chemical synapses between the two ens using analog circuitry . here
we report on results obtained by using a software implementation of the chemical synapses which allows to investigate the role of the synaptic time constant @xmath42 . in the analog circuit implementation of the chemical synapses we need to replace a capacitor every time we want to change the time constant , but in the software version
this time constant is just a parameter , so it is easier to study the role of these time constant in the software version . in this paper
the time constant is fixed , and our observations on the role of a changing time constant will be reported in another paper .
the results using a software version of the chemical synapse , and the results using our hardware version were identical .
we used the nonlinear amplifiers to reshape the signals corresponding to the membrane potential of the ens in such a way that the new signals had amplitudes , spike / burst ratios , and voltage offsets close to the the signals generated by living neurons . with these reshaped signals we used new dynamic clamp software @xcite to generate in real time the currents corresponding to the graded chemical synapses
as described the by first - order kinetics @xmath43 and @xmath44 where @xmath45,\ ] ] when @xmath46
. otherwise @xmath47 .
@xmath48 is the maximal synaptic conductance , @xmath49 , the instantaneous synaptic activation , @xmath50 , the steady - state synaptic activation , @xmath51 , the synaptic reversal potential , and @xmath52 and @xmath53 are the presynaptic and postsynaptic voltages respectively .
@xmath42 is the synaptic time constant , @xmath54 , the synaptic threshold voltage , and @xmath55 , the synaptic slope voltage .
the synaptic reversal potentials were selected so that the currents injected into the postsynaptic ens were always negative for inhibitory synapses and positive for excitatory synapses , emulating the biological synapses @xcite .
the synaptic threshold voltages were set in the middle of the amplitude of the bursts , and the synaptic slope voltage was adjusted to make the output of the hyperbolic tangent slightly saturated at the spikes . in our experiments @xmath48 was varied as we collected different data sets .
we used standard values for the other parameters in the dynamic clamp program : @xmath56 ( inhibitory synapses ) or @xmath57 ( excitatory synapses ) ; @xmath58 ; @xmath59 ; and @xmath60 .
as before we waited for at least 40 seconds after connecting the ens with the chemical synapses before starting the recording of the 20 seconds of data from the membrane potential of the ens .
to analyze the degree of synchronization of slow bursts between two coupled neurons ( electrically or chemically ) we proceed in the same manner as we used for our experiments on synchronized living neurons @xcite .
this was based on a method developed for the experimental studies of synchronization of chaotic oscillations in electronic circuits @xcite .
we used an overlap - add method of fir filtering with a hamming window , and used an fft and a cutoff frequency of 5hz to suppress the spikes , obtaining the filtered data series @xmath61 ; @xmath62 .
the synchronization of the ens is quantified by calculating the difference @xmath63 , and studying the normalized standard deviation @xmath64 and the normalized maximal deviation @xmath65 as a function of @xmath40 for the electrical coupling or as a function of @xmath48 for the chemical coupling . for notational convenience ,
we indicate excitatory couplings with values of @xmath66 and inhibitory couplings with values of @xmath67 .
the parameters of the isolated neurons were set in the chaotic spiking - bursting regime .
an example of the behavior of an isolated en is shown in figure [ free ] .
note that the scale for @xmath15 is double that of the numerical simulations shown in figure [ 4dphases ] because of the rescaling time constant in the analog integrator ( see section [ electronicneurons ] ) .
the relative behavior of the spikes and slow oscillations can be seen in the plots of @xmath68 vs. @xmath69 ( figure [ free]a ) and @xmath70 vs. @xmath71 ( figure [ free]b ) , respectively .
we began with electrical coupling between two four dimensional analog circuit models implementing equation([eq : hrmodel ] ) .
we varied only @xmath72 keeping all other parameters fixed .
a convenient representation of the range of behavior we observed is presented in figure [ mapel ] . here ,
overlying values of @xmath73 and @xmath74 we give a verbal description of the quantitative behavior of time series in each regime .
to illustrate the phenomena seen in each regime @xmath75 $ ] of figure [ mapel ] we show examples of the time series for the membrane potentials @xmath76 of the two neurons in figures [ posel ] and [ negel ] .
* when @xmath78 the two neurons are uncoupled and display independent chaotic oscillations as shown in figure [ free ] .
* for small , positive coupling @xmath79 , regions of nearly independent chaotic spiking - bursting activity are observed as well as some regions of synchronized bursting activity as shown in figure [ posel](b ) where we set @xmath80 .
there is a small range of @xmath81 @xmath82 in which intermittent anti - phase bursting behavior can be found .
the burst length in this case is kept nearly regular from burst to burst as shown in figure [ posel](a ) . * for @xmath83 the behavior is still chaotic for the two neurons but most of the bursts are synchronized as shown in figure [ posel](c ) where we set @xmath84 . * from @xmath85 the bursting activity becomes regular going from a region in which there is partial synchronization ( spikes not synchronized ) , as shown in figure [ posel](d ) where we set @xmath86 , to a region of total synchronization ( bursts and spikes synchronized ) , shown in figure [ posel](e ) where we set @xmath87 . * from @xmath88 there is total synchronization in the spiking - bursting activity , and the oscillations are chaotic as shown in figure [ posel](f ) where we set @xmath89 .
+ results for @xmath90 * for negative coupling @xmath91 , the oscillations are predominantly chaotic and the hyperpolarizing regions , where the membrane voltage is quite negative , of the signals are all in anti - phase .
the average burst length decreases as the coupling becomes stronger as shown in figure [ negel ] . for a small range of @xmath40 @xmath92 very long bursts
were observed as shown in figure [ negel](a ) .
@xmath73 and @xmath74 provide quantitative measures of the synchronization between two ens . in our report on the experimental work @xcite with two biological cells ,
the results for @xmath73 and @xmath93 can be seen in figure 5 of that paper .
note that , as in the case of coupled biological neurons , we have here a bifurcation between positive and negative electrical coupling . in the experimental work on electrically coupled biological neurons a value for the external coupling @xmath94ns serves to null out the natural coupling of about that amount , so the figures here and in the earlier paper are to be compared by sliding @xmath95 here to @xmath96ns there . both in the biological and electronic experiments , the sharp phase transition from very small @xmath97 for positive coupling to large , nearly constant values is associated with the rather rapid change from nearly and then fully synchronous behavior for positive couplings to out - of - phase oscillations for negative couplings . the @xmath73 and @xmath74 curves in the paper on coupled biological neurons @xcite shows far fewer points and consequently less detail that our curves for coupled 4d ens .
clearly this is because of the resolution in the biological experiments and the difficulty in performing experiments at such closely chosen values of @xmath98 . at this time
the details of behavior revealed in the present experiments on ens have not been verified in the biological setting .
one should view our figure [ mapel ] and figure 5 of @xcite as in excellent qualitative agreement .
we have observed the behavior of two four dimensional ens coupled with _
identical _ chemical synapses .
two electrical versions of chemical synapses were built with identical parameters and then used to coupled two four dimensional ens . in the equations ,
equation ( [ chemsynapse ] ) , we integrate to represent the chemical synapse , all parameters were set equal in the two connecting circuits .
we then varied @xmath99 in each chemical synapse over the range @xmath100 for an excitatory synapse , namely @xmath57 , and over the range @xmath101 for an inhibitory synapse , namely , @xmath102 .
the other parameters were held fixed at @xmath103 , and @xmath60 . in figure [ mapchem ]
we collect the statistical results , expressed in our usual quantities @xmath104 and @xmath105 , for both excitatory and inhibitory synaptic connections .
negative values of @xmath99 represent inhibitory connections .
this , perhaps apparently peculiar , method of presentation allows us to see immediately the relationship between excitatory and inhibitory interconnections . as earlier with electrical couplings we provide a verbal description of each region of behavior over the whole range of @xmath99 .
we show examples of the time series for the membrane potential @xmath15 of the two neurons in figure [ exchem ] and figure [ inchem ] . when coupled with implementations of excitatory chemical
synapses the ens displayed the following behaviors : * when @xmath106 the two neurons are uncoupled and display independent chaotic oscillations as shown in figure [ free ] . * for positive coupling @xmath107 a transition from the chaotic behavior to regular spiking / bursting
is observed . for small coupling the independent chaotic spiking / bursting activity of the uncoupled neurons
is replaced by a behavior in which most of the bursts are synchronized , but the oscillations are still chaotic as shown in figure [ exchem](a ) for @xmath108 .
as @xmath48 is increased all the bursts become synchronized , and the activity becomes periodic as shown in figure [ exchem](b ) for @xmath109 . * for @xmath110 the bursts remain synchronized and get longer , but there are no longer any spikes during the bursts as shown in figure [ exchem](c ) for @xmath111 .
finally we report on our experiments with an electronic version of an inhibitory chemical synapse .
this inhibitory synaptic coupling occurs in the lobster pyloric cpg as well as many other cpgs , and we have suggested @xcite that inhibitory chemical coupling will lead to regularization of the chaotic oscillations of the individual neurons .
* for small @xmath48 the oscillations are still chaotic , but all of the hyperpolarizing regions of the membrane voltages are in anti - phase as shown in figure [ inchem](a ) for @xmath112 .
* when @xmath113 the oscillations become periodic , and all the hyperpolarizing regions are in out - of - phase as shown in figure [ inchem](b ) .
* for @xmath114 the out - of - phase behavior of the hyperpolarizing regions remains , but the oscillations are chaotic again as shown in figure [ inchem](c ) for @xmath115 . * for @xmath116 the oscillations regularize again , and the behavior is periodic with out - of - phase bursting as shown in figure [ inchem](d ) for @xmath117 and in figure [ inchem](e ) for @xmath118 . * for @xmath119 the oscillations are chaotic and long out - of - phase bursts are observed as shown in figure [ inchem](f ) for @xmath120 .
the only experiments we know which relate to these observations on two chemically coupled ens are not a precise match , but bear noting .
r. elson @xcite has isolated a pair of lp and pd neurons from the lobster pyloric cpg ; these have mutual inhibitory coupling .
elson varied the strength of the chemical coupling using neuromodulators and making measurements at four values of @xmath99 over a nominal rage of 20 ns to 60 ns .
he observed only the behavior reported in the penultimate item of our experiments on inhibitory coupling .
unfortunately , control of the identity of the mutual inhibitory couplings was not possible , nor was it possible for us to directly compare the calibration of elson s indication of the magnitude of @xmath99 with our own choices in using ens .
to date then , we have no direct laboratory evidence on synchronization of biological neurons mutually coupled with chemical synapses .
this is in contrast to our observations on electrically coupled biological neurons @xcite .
this represents an interesting opportunity for biological experiments which may be directly compared to our results using ens .
we have previously reported experiments on replacing the ab neuron from the pyloric cpg in its interaction with an isolated pair of pd neurons with a * three dimensional * en @xcite .
for completeness in light of the work reported in this paper , we carried out an experiment in which one of our four dimensional neurons was coupled bidirectionally to one of the pd neurons in the ab / pd pacemaker group of the pyloric cpg .
the full description of the methods used in the biological preparation will appear elsewhere @xcite , but here we quite briefly summarize those points important to the main thrusts of this article .
these experiments were carried out on one of the two pyloric dilator ( pd ) neurons from the pyloric central pattern generator ( cpg ) of the lobster stomatogastric ganglion ( stg ) @xcite .
the stg of the california spiny lobster , _ panulirus interruptus , _ was removed using standard procedures and pinned out in a dish lined with silicone elastomer and filled with normal lobster saline .
the stg was isolated from its associated anterior ganglia , which provide activating inputs , by cutting the stomatogastric nerve .
two glass microelectrodes were inserted in the soma of the pd neuron : one for intracellular voltage recording and another one for current injection .
the voltage signals were digitized at 10000 samples / sec .
the two pd neurons remained coupled to each other and to the autonomously bursting interneuron ( ab ) by their natural electrical synapses , but were isolated from the rest of the cpg by blocking chemical input synapses with picrotoxin @xmath121 .
the artificial electrical coupling was provided by injecting in the en and in the pd opposite currents .
more details of the experimental setup can be found in @xcite .
the membrane voltage of the en was reshaped to make its amplitude ratio in spiking / bursting mode , its total amplitude and its voltage offset similar to those of the pd neuron . only electrical coupling , positive and negative ,
is reported here .
we connected the neurons with the analog electrical synapse and observed their spiking - bursting behavior as shown in figure [ hybrid ] . when uncoupled , the neurons had independent spiking / bursting activity as shown in figure [ hybrid](b ) .
for large enough negative coupling the neurons are synchronized and fire out - of - phase as shown in figure [ hybrid](a ) . for positive coupling the neurons show synchronized bursting activity as shown in figure [ hybrid](c ) . for this value of @xmath98
the bursts are synchronized but not the spikes .
this result is in agreement with the experiments made with a pair of electrically coupled ens , as we discussed above , as well as for a pair of living stg neurons @xcite .
the ens described in this paper are simple analog circuits which integrate four dimensional differential equations representing fast and slow subcellular processes that give rise to the characteristic spiking and spiking - bursting behavior of cpg neurons .
the single neurons can be easily set into a chaotic regime that reproduces the irregular firing patterns observed in biological neurons when isolated from the rest of the cpg .
this study comprises : ( a ) two electrically coupled ens and ( b ) two ens connected with excitatory and inhibitory chemical synapses .
these two types of connections exist in almost all known cpgs .
the range of observations summarized in figures [ mapel ] and [ mapchem ] shows the rich behavior and complexity of these minimal network configurations .
it indicates how small changes in the coupling conductance can drive the cells into completely different regimes .
in particular , some of our experiments predict the appearance of chaotic out - of - phase synchronization for different coupling configurations .
these results are displayed in figures [ negel]c and [ inchem]f . in general , the experiments with the ens contribute directly to our understanding of the origin of regularization of individually chaotic neurons through cooperative activity .
how complicated should one require a model neuron to be ?
in our view the answer depends on the neural function one wishes to represent .
the analysis of the electrical activity of isolated neurons from the lobster pyloric cpg indicates that the number of active degrees of freedom is not very large , ranging from three to five in various environments , and this suggests a very simple representation in terms of dynamical equations .
our analysis @xcite of much richer hodgkin - huxley models of these individual neural oscillators also indicates that in the regime of biological operation , the number of active degrees of freedom is equally small . on this basis
we developed the hindmarsh - rose type models of these neurons both in numerical simulation and in analog electrical circuitry .
this paper has moved that inquiry about the complexity of representation for the components of a biologically realistic neural network to another level .
here we have investigated whether the simplified neural models , when coupled together in small networks , but in biologically realistic manners , can reproduce our observations biological neurons alone . the striking result of the observations presented here , when the experimental setup matches that of the biological networks , is that the observed behavior of the ens also matches .
further , using our ens , we are able to make distinct predictions about the behavior of biological or hybrid ( biological and en ) networks in settings not yet investigated .
our experiments on coupled biological neurons and ens provide further ground for testing the validity of numerical and electronic models of individual neural behavior as well as presenting interesting new examples of coupled nonlinear oscillators .
hybrid circuits with biological and electronic neurons coupled together are a powerful mechanism to understand the modes of operation of cpgs .
the hybrid system constitutes an easy way to change the connectivity and global topology of the cpg .
the roles of intrinsic dynamics of the neurons and the synaptic properties of the network in rhythm generation can be easily studied with these hybrid configurations @xcite ) .
there are previous efforts studying electronic neurons alone and in conjunction with biological neurons .
an early example is the work of yarom @xcite where a network of four oscillators , realized as an analog circuit , was interfaced with an olivary neuron in a slice preparation .
yarom studied the response of the olivary neuron when it received oscillating electrical input from the network .
there was no feedback from the biological neuron to the network he constructed .
le masson , et al @xcite developed a digital version of a neuron comprising a hodgkin - huxley @xcite ( hh ) model of various pyloric cpg neurons with three compartments and eight different ion channels which ran on a dsp board located on the bus of a personal computer .
they connected this model into a variety of different configurations of subcircuits of the pyloric cpg replacing at various times the lp , a pd or a py neuron .
using this ` hybrid ' setup they verified that many aspects of the pyloric rhythm are accurately reproduced when their dsp based neuron replaces one of the biological neurons in their system . in subsequent work
@xcite , this group has developed vlsi devices for integrating the hh models and has utilized them in mixed circuits ( ens and biological neurons ) , replacing the dsp version of the conductance models in their biological preparations .
the complexity of these ens has not been needed in our modeling nor in the further experiments on their interaction with each other as reported here .
we have not found any reports in the literature on the mutual interaction of these analog vlsi neural circuits .
there are two interesting directions to which the results reported here may point : * \(1 ) biologically realistic neural networks of much greater size than the elementary ones investigated here may be efficiently investigated numerically or in analog circuitry using the realistic , but simple hr type models .
the integration of the model equations is no challenge to easily available computing power and large networks should be amenable to investigation and analysis . * \(2 ) the networks investigated here are subcircuits of a biological circuit of about fifteen neurons which has the functional role of a control system : commands are presented from other ganglia of the lobster and this pyloric circuit must express voltage activity to the muscles to operate a pump for shredded food passing from the stomach to the digestive system .
many other functions are asked of biological neural networks . using the full richness of hh models for
the component neurons may seem attractive at one level , but the results presented here suggest that many interesting questions may be asked of those networks using the simplified component neurons studied here .
r.d . pinto was supported by the brazilian agency fundao de amparo pesquisa do estado de so paulo - fapesp .
pv acknowledges support from mec . partial support for this work came from the u.s department of energy , office of science , under grants de - fg03 - 90er14138 and de - fg03 - 96er14592 .
we also acknowledge the many conversations we have had with ramon huerta , rob elson and allen selverston on the dynamics of cpg neurons .
renaud - le masson , s. , e. marder , g. le masson , and l. abbott , `` hybrid circuits of interacting computer models and biological neurons '' , _ neural information processing systems 5 _ , pp .
813 - 819 san mateo , ca .
morgan kaufmann publishers , ( 1993 ) .
le masson , g. , s. le masson , and m. moulins , `` from conductances to neural network properties : analysis of simple circuits using the hybrid network method '' , _ prog .
biophys . molec .
biol . _ * 64 * , 201 - 220 ( 1995 ) .
laflaquiere , a. , le masson , s. , dom , j.p . , and le masson , g. , `` accurate analog vlsi model of calcium - dependent bursting neurons '' , ( 1997 ieee international conference on neural networks .
proceedings ( cat .
no.97ch36109 ) , proceedings of international conference on neural networks ( icnn97 ) , houston , tx , usa , 9 - 12 june 1997 . )
new york , ny , usa : ieee , 1997 . p.882 - 7 vol.2 . 4 vol .
xlvi+2570 pp . ;
dupeyron , d. , le masson , s. , deval , y. , le masson , g. , dom , j .-
, `` a bicmos implementation of the hodgkin - huxley formalism , '' ( proceedings of the fifth international conference on microelectronics for neural networks and fuzzy systems .
microneuro96 , proceedings of fifth international conference on microelectronics for neural networks , lausanne , switzerland , 12 - 14 feb .
los alamitos , ca , usa : ieee comput .
soc . press , 1996 .
p.311 - 16 .
x+370 pp ;
le masson , s. , deval , y. , le masson , g. , tomas , j. , dupeyron , d. , dom , j.p , `` a bicmos asic for modeling biological neurons '' , ( proceedings of the fourth international conference on microelectronics for neural networks and fuzzy systems , proceedings of the fourth international conference on microelectronics for neural networks and fuzzy systems , turin , italy , 26 - 28 sept .
los alamitos , ca , usa : ieee comput .
soc . press , 1994 .
p.202 - 6 .
xi+457 pp .
an excellent introduction to the issues involved , the approximations employed , the methods of implementation as well as to the results of some experiments can be found in the dissertation of g. le masson available online at . | we report on experimental studies of synchronization phenomena in a pair of analog electronic neurons ( ens ) .
the ens were designed to reproduce the observed membrane voltage oscillations of isolated biological neurons from the stomatogastric ganglion of the california spiny lobster _ panulirus interruptus_. the ens are simple analog circuits which integrate four dimensional differential equations representing fast and slow subcellular mechanisms that produce the characteristic regular / chaotic spiking - bursting behavior of these cells . in this paper
we study their dynamical behavior as we couple them in the same configurations as we have done for their counterpart biological neurons .
the interconnections we use for these neural oscillators are both direct electrical connections and excitatory and inhibitory chemical connections : each realized by analog circuitry and suggested by biological examples .
we provide here quantitative evidence that the ens and the biological neurons behave similarly when coupled in the same manner .
they each display well defined bifurcations in their mutual synchronization and regularization .
we report briefly on an experiment on coupled biological neurons and four dimensional ens which provides further ground for testing the validity of our numerical and electronic models of individual neural behavior .
our experiments as a whole present interesting new examples of regularization and synchronization in coupled nonlinear oscillators .
= 6.6 in = 9.10 in = -0.0 in = -0.5 in |
the most luminous known agn with an ionized absorber is the radio - loud lobe - dominated quasar 3c351 ( l@xmath5 erg s@xmath6 , z=0.371 , fiore et al . , 1993 ) .
ionized absorbers are common in low redshift , low luminosity seyfert 1 galaxies ( reynolds , 1997 ) , but rare in higher redshift , higher luminosity quasars .
high luminosity agns are very likely physically larger and so may exhibit slower time variability and have different physical conditions .
3c 351 has a very high uv to x - ray ratio ( @xmath7 , tanambaun et al . , 1989 ) and
its ir to uv spectrum does not show any evidence of reddening , in contrast to several low redshift seyfert 1 galaxies thought to host dusty / warm absorbers .
only two other radio - loud quasars , 3c 212 ( mathur et al . , 1994 ) and 3c 273 ( grandi et al , 1997 ) , have candidate ionized absorbers .
the main absorption feature in the rosat pspc x - ray spectrum of 3c 351 is a deep edge at @xmath1 kev ( quasar frame ) , which is likely to be due to ovii - oviii .
mathur et al . ( 1994 ) built a simple one - zone model of this absorber that also explained the ovi , civ and ly@xmath8 uv absorption lines seen in a nearly simultaneous hst fos spectrum .
such simple models , although elegant , have come under criticism .
ionized absorbers in some seyfert galaxies show different variability behavior in different absorption features , some of which are not as predicted for gas in photoionization equilibrium .
multiple absorbing zones have been introduced to explain these effects ( e.g. reynolds 1997 ) . in another paper ( nicastro et al . , 1998 )
we instead explore additional physics ( non - equilibrium photonionization models and collisional models ) while retaining a single zone of absorbing gas .
3c 351 is a relatively slowly varying source , compared to the rapidly variable seyfert with warm absorbers .
this persistent variability on timescales that span possible ionization and recombination times makes it hard to define an initial equilibrium state . in contrast 3c 351
presents a simplified situation in which to study changes in the ionization state in response to luminosity changes . here
we show that in 3c 351 at least the simplest photoionization equilibrium , single zone model continues to be sufficient .
3c351 was observed twice by rosat , on 1991 october and 1993 august .
the first observation was reported by fiore et al .
the second observation was then proposed in order to search for time variability that could test photoionization models .
fortunately a factor 1.7 decrease in the pspc count rate was seen , providing just such a test . here
we present the second data set and compare the results with predictions .
we considered the two pspc observations of 3c351 taken roughly two years apart , in 1991 october and 1993 august .
table 1 gives , for each observation , the observation date , its duration , the net exposure time , the net source counts , the count rate , and the signal to noise ratio .
the data reduction and the timing analysis were performed using xselect .
we reduced the second pspc observation of 3c351 following the procedures used in fiore et al .
( 1993 ) for the first observation .
for the first observation we used the spectrum and light curve obtained by fiore et al .
the source intensity was consistent with a constant value during both pspc observations , with no more than 20% change in the @xmath9 day and @xmath10 day spans of the two observations .
3c 351 thus appears more stable than the lower luminosity seyferts with ionized absorber ( reynolds , 1997 ) . since 3c 351
has a flat x - ray spectrum and broad optical emission lines it is consistent with the trend exhibited by the three similar but radio - quiet , pg quasars of the fiore et al .
( 1998 ) sample .
those quasars show little or no variability on timescales as short as 10 days ( compared to the narrow line objects of that sample which are rapidly variable and have steep x - ray spectra ) .
however the mean count rate dropped by a factor @xmath11 during the 22 months between the observations ( see table 1 ) , again consistently with the fiore et al .
( 1998 ) discovery .
hereinafter we shall call these the high and low states , respectively . [ cols="^,^,^,^,^,^,^,^ " , ] @xmath12 in @xmath13 ph s@xmath6 @xmath14 kev@xmath6 ( at 1 kev ) .
= 4.5 in = 3.in collisional ionization models can fit the data equally well , as they should at low column densities ( nicastro et al . ,
1997 ) . to do
so however they require arbitrary changes , by a factor 1.5 in temperature or a factor 2 in n@xmath15 .
instead photoionization equilibrium models predict the observed correlation with the ionizing continuum . to the authors
this is a strong argument in favor of a photoionization model .
the behavior of the main physical properties of the absorber can be seen in a color - color diagram . in fig .
3 we plot the hardness ratios hr = h / m against the softness ratio sr = s / m from the count rates in three bands ( s=0.15 - 0.58 kev , m=0.88 - 1.47 kev , and h=1.69 - 3.40 kev , at z=0.371 ) for theoretical curves ( for log(@xmath16)=21 , 21.5 , 22 and 22.14 ) obtained by folding the equilibrium photoionization models ( for log(u ) in the range -0.3 - 1.5 , and galactic @xmath16 ) with the response matrix of the pspc .
= 4.0 in = 4.0 in the position of a point in this diagram readily gives the dominant ion in the gas .
the rapid change of sr as u increase from 0.5 to 5.5 ( on the curve corresponding to log(@xmath16)=22.14 ) corresponds to large increases of the transparency of the gas at e@xmath17 kev as h and he becomes rapidly fully ionized .
as u further increases from 5.5 to 16.5 the sr color changes more slowly , and inverts its trend at u@xmath18 , with hr now changing more rapidly than sr .
the inversion point indicates the switch from an ionization state dominated by ovii to that dominated by oix . in the last part of the curve as u increases , sr and hr decrease until all the ions in the gas are fully stripped and the gas is completely transparent to radiation of any energy .
the two data points show the two 3c 351 observations .
the best fit u in the high observation is marked on the log(@xmath16)=22.14 curve along with the value obtained by scaling by the intensity ratio between the two observations ( a factor 1.7 ) .
the two points are consistent with the position of the observed colors of 3c351 in the two observations , so the predictions of the simple equilibrium photoionization model is consistent with the data .
the as predicted change of the ionization parameter in the 3c 351 absorber to a change in the ionizing continuum is strong evidence that photoionization is the dominant ionization mechanism .
we also see that the absorber comes to ionization equilibrium within 22 months , and can use this to constrain the physical properties of the absorber . the time @xmath19 measures the time the gas needs to reach equilibrium with the instantaneous ionizing flux ( nicastro et al , 1997 ) .
this time depends on the particular ionic species considered .
the ionic abundances of oxygen in the absorber in 3c 351 are distributed mainly between only two ionic species : ovii and oviii . in this simple case
a useful analytical approximation for @xmath19 is : @xmath20_{t+\delta t}}}\nonumber \normalsize\end{aligned}\ ] ] where _ eq _ indicates the equilibrium quantities . by requiring
that @xmath19 for ovii and oviii species is shorter than the @xmath21 s ( quasar frame ) elapsed between the two observations , we can find a lower limit to the electron density of the absorber : @xmath22 @xmath3 .
since we know the values of u and @xmath23 , this density limit translates into a limit on the distance of the ionized gas from the central source , r@xmath24 pc .
these limits are consistent with the upper limit of @xmath25 @xmath3 , and r@xmath26 pc , found by mathur et al .
( 1994 ) using a lower limit for the distance of the cloud from the central source , based on the absorber being outside the broad emission line region .
the relative closeness of these two limits ( factor @xmath0100 ) implies that a variation in shorter times ( @xmath27 month ) , would be likely to show non - equilibrium effects ( nicastro et al , 1997 ) .
if a factor 2 flux change in @xmath28 week showed no such effects the simplest photoionization model would have to be abandoned .
we have tested ionization models for the ionized absorber in 3c 351 .
in particular , we tested a simple one zone photoionization equilibrium model on two pspc spectra of 3c 351 that show a factor @xmath29 decrease in flux .
the model correctly predicts the sense and amplitude of the observed change in the ionization state of the absorber , correlated with the ionizing continuum flux .
given that photoionization equilibrium applies we can derive a lower limit to the electron density of the absorber : @xmath30 @xmath3 .
this is consistent with the upper limit of @xmath31 @xmath3 found by mathur et al .
the distance of the ionized gas from the central source is then 0.3 pc @xmath4 r @xmath4 50 pc .
the closeness of these two limits creates a strong test of photoionization models : factor 2 variations in 3c 351 on timescales of order a week _ must _ show non - equilibrium effects . | we present two rosat pspc observations of the radio - loud , lobe - dominated quasar 3c 351 , which shows an ` ionized absorber ' in its x - ray spectrum .
the factor 1.7 change in flux in the @xmath02 years between the observations allows a test of models for this ionized absorber .
the absorption feature at @xmath1 kev ( quasar frame ) is present in both spectra but with a lower optical depth when the source intensity - and hence the ionizing flux at the absorber - is higher , in accordance with a simple , single - zone , equilibrium photoionization model .
detailed modeling confirms this agrement quantitatively .
the maximum response time of 2 years allows us to limit the gas density : @xmath2 @xmath3 ; and the distance of the ionized gas from the central source r @xmath4 19 pc .
this produces a strong test for a photoionized absorber in 3c 351 : a factor 2 flux change in @xmath01 week in this source _ must _ show non - equilibrium effects in the ionized absorber . |
in recent years , several research groups @xcite are involved in the investigations of nanoparticle spinel oxides because of their potential applications in magnetic devices , in micro wave technology @xcite , in high density magnetic recording media @xcite , in magnetic fluids as drug carrier etc .
@xcite . various types of nanoparticle materials such as , metal : fe , co , ni @xcite , metallic alloys : fe - cu @xcite , and metallic oxides : mnfe@xmath2o@xmath3 @xcite and znfe@xmath2o@xmath3 @xcite , are under current research activity .
while metal and inter metallic nanoparticles suffer from stability problems in atmospheric condition , metallic oxides are highly stable under ambient conditions @xcite . various factors such as ,
particle size distribution @xcite , inter - particle interactions @xcite , grain ( core ) and grain boundary ( shell ) structure @xcite and metastable structure of the system @xcite control the properties of nanoparticles .
some of the specific properties of the nanoparticles which are of interest are quantum magnetic tunneling @xcite , various magnetic order like ferrimagnet / ferromagnet , spin glass / superparamagnet and spin canting effects @xcite . + the interesting aspect of magnetism in spinel oxides is that the magnetic order is strongly dependent on the competition between various superexchange type interactions i.e. , j@xmath7 ( a - o - b ) and j@xmath8 ( b - o - b ) , where a : tetrahedral ( a ) sites moments and b : octahedral ( b ) sites moments , o : is o@xmath9 ions @xcite .
certain amount of site disorder _
i.e. _ , the cations redistribution between a and b sites is sufficient to change the super - exchange interactions in nanoparticles spinel @xcite .
hence , the magnetic properties of the nanoparticle spinels can be drastically different from their bulk counterpart . in this light
, we investigated co@xmath0zn@xmath1fe@xmath2o@xmath3 spinel oxide system .
+ the cation distribution of the bulk sample is ( zn@xmath10fe@xmath11)@xmath12[co@xmath13fe@xmath14@xmath15o@xmath3 ( a= tetrahedral sites , b= octahedral sites ) @xcite .
the a site is strongly diluted with respect to b site and b site moments contribute to different size clusters .
the spins inside the clusters form canted structure due to short range ferrimagnetic interactions .
the total magnetization m (= m@xmath15cos@xmath16-m@xmath12 ) of such type of canted system depends not only on the sublattice ( site ) magnetization m@xmath15 and m@xmath12 but also on the canting angle @xmath16 between the b site spins .
the b site spin canting is strongly related with the amount of a site moments .
if more number of a site moments exist in as milled co@xmath0zn@xmath1fe@xmath2o@xmath3 sample , the canting angle between b site spins is expected to be reduced and a drastic change in cluster glass properties of bulk sample @xcite is expected .
unfortunately , most of the reports in literature concern only on the long range order system , such as znfe@xmath2o@xmath3 @xcite and nife@xmath2o@xmath3 @xcite , and the magnitudes of a and b site moments but not on the short range interacting cluster glass system where change of spin canting inside the clusters plays an important role @xcite . in this presentation , we show that for as milled co@xmath0zn@xmath1fe@xmath2o@xmath3 sample , the magnetization enhancement is related with the reduction of b site spin canting and an enhancement of ordering temperature occurs which is related with the cation redistribution and strain induced anisotropy .
stoichiometric amounts of zno ( 99.998% from johnson matthey ) , fe@xmath2o@xmath17 ( 99.998% from johnson matthey ) and co@xmath17o@xmath3 ( 99.5% from fluka ) oxides for the composition co@xmath0zn@xmath1fe@xmath2o@xmath3 were mixed and ground for 2 hours .
then , the mixture was pelletized and sintered at 950@xmath18c for 12 hours and slowly cooled to room temperature .
the pellet was again ground and pelletized and finally sintered at 1000@xmath18c for 12 hours .
the heat treatment was carried out under atmospheric condition with heating and cooling rate of 2 - 3 @xmath18c / minute .
the xrd spectra of the as prepared bulk sample ( s@xmath19 ) confirm formation of well crystalline cubic spinel phase .
the as prepared sample was milled for 24 hours in a spex 8000 mixer / mill using a set of six balls , two of diameter 1/2 inch and four of diameter 1/4 inch with ball to powder weight ratio 5:1 .
the as milled sample ( s@xmath20 ) was heat treated in air at 300@xmath6c for various durations , ranging from 2 hours to 575 hours , to study the temporal phase evolution .
the 6 hours heat treated as milled samples are denoted as s@xmath21 ( for 300@xmath18c ) , s@xmath22 ( for 600@xmath6c ) and s@xmath23 ( for 1000@xmath6c ) .
the heating and cooling rates were maintained at 2 - 3@xmath18c / minute during the heat treatment process .
the x ray diffraction ( xrd ) data ( fig .
1a ) were taken using philips pw1710 diffractometer with cuk@xmath24 radiation in the 2@xmath16 range 10@xmath6 - 90@xmath6 with a step size 0.02@xmath18 .
the xrd data of as prepared bulk sample show narrow crystalline lines with estimated particle size of a few micrometers .
after 24 hours milling of the bulk sample , the xrd peak lines become significantly broad and no other phases are seen , except the cubic spinel structure .
the broadening in xrd lines suggest nano structure or non - uniform micro strain introduced in the lattices during milling process @xcite .
the width of @xmath25 xrd peak line gives average particle size @xmath4 20 nm for the as milled sample ( s@xmath20 ) .
when the as milled sample is heat treated , the xrd lines ( fig.1b - c ) become sharp , which indicates recrystallization process in the system . the average particle size increases with annealing temperature as 23 nm , 28 nm and 62 nm for the samples s@xmath21 , s@xmath22 and s@xmath23 , respectively .
eventhough the xrd spectrum of the s@xmath23 sample is very similar to that of s@xmath19 sample , the average particle size @xmath4 62 nm suggest that complete restoration of crystallinity close to s@xmath19 sample is yet to be reached .
we have determined the chemical composition of s@xmath20 sample by x - ray fluorescence ( xrf ) technique .
the fluorescence spectrum was detected using ortec hpge detector and finally recorded in a pc based multichhanel analyzer . the xrf spectrum ( fig .
2 ) shows the characteristic k@xmath24 and k@xmath26 lines of fe , co and zn elements .
no extra lines have been found that correspond contamination from hardened steel balls with composition fe@xmath27cr@xmath28ni@xmath29 .
the peak integrals of the k@xmath24 lines of all the elements are obtained using the standard peak fitting programme to get the relative percentage .
this program uses the principle of fundamental parameter method @xcite .
the obtained composition is : fe = 62.45% , co = 7.17% and zn = 30.38% which is very close to the expected values _
i.e. _ fe = 63.53% , co = 6.71% and zn = 29.76% for co@xmath0zn@xmath1fe@xmath2o@xmath3 system .
the maximum error of the fitted values are within 5% .
the low field dc magnetization ( t = 20k to 320k , h= 10 oe to 100 oe ) and ac susceptibility ( t= 60k to 330k , h@xmath30 @xmath4 1 oe and frequency ( f ) = 337 hz and 7 khz ) measurements have been performed using home made magnetometer @xcite . in zfc condition , the sample has been cooled from room temperature to 20k in the absence of dc magnetic field , then the field was applied at 20k and magnetization data were recorded while increasing the temperature . in fc condition ,
the sample was cooled from room temperature to 20k in presence of dc magnetic field ( same magnitude which was applied during zfc measurement ) and the magnetization data were recorded with increasing temperature and keeping the field on .
high field magnetization , hysteresis experiments were performed using vsm magnetometer in the fields upto @xmath3112 tesla in the temperature range 10k to 300k .
+ mssbauer spectra were recorded at 300k without applying external magnetic field , using a constant acceleration spectrometer in transmission geometry mode .
the spectra were recorded using a 5 mci @xmath32co in rh matrix source .
the hyperfine magnetic field ( hmf ) distribution at the @xmath32fe nuclei was evaluated from the mssbauer spectra using the method of le car and dubois @xcite . in this model
a linear relationship between hmf ( h ) and isomer shift ( is ) is assumed in the form is = _
a_h + _ b _ , where _ a _ and _ b _ are the fitting parameters to get a minimum @xmath33 .
the isomer shift ( is ) wes calculated with respect to @xmath32fe metal .
3 shows the real ( @xmath34 ) and imaginary ( @xmath35 ) components of ac susceptibility for sample s@xmath20 .
it shows a broad @xmath34 maximum at t@xmath15 @xmath4 320k ( for h@xmath30 @xmath4 1 oe , f= 337 hz ) and the corresponding @xmath35 maximum at @xmath4 290k .
the broadness of the ac susceptibility maxima suggest a cluster size distribution in the as milled sample s@xmath20 .
the temperature where @xmath34 shows maximum is defined as the average blocking temperature ( t@xmath15 ) of the clusters . from the frequency dependence of ac susceptibility measurements ( fig .
3a ) , a positive frequency shift of t@xmath15 is observed in the @xmath34 as well as in the @xmath35 maximum in the measurement frequency range .
this is a characteristic of superparamagnetic blocking of the ferromagnetic clusters in different metastable states @xcite . in the low temperature region ,
the @xmath34 data show a small shoulder at @xmath5 100k ( designated as t@xmath36 ) and @xmath35 data show ( fig .
3b ) increasing tendency below t@xmath36 .
this type of behaviour of t@xmath36 can be attributed due to the disordered surface or grain boundary spins in a nanoparticle system @xcite .
+ in order to check the effect of heat treatment on magnetic properties , we have carried out the ac susceptibility measurements for the s@xmath23 sample .
this heat treated sample shows mixed magnetic phases .
the @xmath34 and @xmath35 data of s@xmath23 sample ( fig .
4 ) show two magnetic ordering at t@xmath37 @xmath4 70k and t@xmath38 @xmath4 280k respectively .
the frequency shift of t@xmath39 follows vogel - fulcher law @xmath40 with characteristic frequency @xmath41 @xmath4 10@xmath42 hz , activation energy e@xmath43 @xmath4 379 ev and constant t@xmath19 @xmath4 47k .
the corresponding frequency shift per decade of _ f _ ( @xmath44/t@xmath45 ) is @xmath4 0.05 , which characterizes spin glass like ordering at t@xmath37 @xcite .
the ordering temperature at t@xmath38 does not show significant frequency shift ( not shown in fig .
4 ) , suggest ferrimagnetic ordering temperature at t@xmath38 @xmath4 280k .
the spin glass transition temperature t@xmath37 @xmath4 70k and ferrimagnetic ordering temperature at t@xmath38 @xmath4 280k of s@xmath23 sample are slightly different with respect to the cluster spin glass freezing temperature t@xmath37 @xmath4 110k and short range ferrimagnetic ordering temperature t@xmath46 @xmath4 260k for bulk sample @xcite .
but it is clear that the mixture of two magnetic phases ( spin glass / cluster spin glass plus ferrimagnetic phase ) of s@xmath23 sample is very similar to the bulk sample s@xmath19 .
the zfc magnetization of s@xmath20 sample ( fig .
5 ) shows a broad maximum at the cluster blocking temperature t@xmath15 @xmath4 320k ( h @xmath5 30 oe ) with a thermomagnetic irreversibility between zfc and fc magnetization below t@xmath15 .
the decrease of zfc magnetization below t@xmath15 is due to the blocking of ferromagnetic clusters in different metastable states , whereas the fc magnetization increases below t@xmath15 due to the orientation of these ferromagnetic clusters in metastable states which give rise to more magnetic contribution @xcite .
it is observed ( fig.5a and fig .
5b ) that the t@xmath15 is highly applied field dependent ( _ i.e. _ , t@xmath15 @xmath4 320k , 280k and 50k for applied field 30 oe , 100 oe and 1 tesla , respectively ) as expected for superparamagnetic blocking of the clusters @xcite . the superparamagnetic behaviour of the as milled sample ( s@xmath20 ) can be further identified from the temperature dependence of field cooled thermoremanent magnetization ( trm ) data ( fig .
5b inset ) that the trm value reduces to zero at t@xmath4 t@xmath15 .
the zero trm value indicate that the effective inter - cluster interactions are also negligible and the clusters behave as a non - interacting small particle above t@xmath15 @xmath4 320k .
+ an interesting feature is observed when we compare the magnetization data at 100 oe of as prepared bulk sample , as milled sample and heat treated samples ( fig .
the magnetization is enhanced when the bulk sample is milled for 24 hours . on heat treating the as milled sample , it is observed that magnetization further increases for s@xmath21 sample
. then magnetization decreases for the s@xmath22 sample .
if the as milled sample is heat treated at higher temperature ( s@xmath23 ) , the magnetization increases . but compared to the bulk sample s@xmath19 , s@xmath23 sample has a higher magnetization value at low temperature and lower value at temperatures @xmath47 200k .
similar non - equilibrium magnetic behaviour as a function of annealing temperature has been observed for nife@xmath2o@xmath3 @xcite .
the superparamagnetic blocking temperature t@xmath48 ( indicated in fig .
6 ) decreases to 270k and 120k for s@xmath21 and s@xmath22 , respectively , in comparison with t@xmath48 @xmath4 280k for s@xmath20 sample at h= 100 oe .
the 100 oe magnetization data of the s@xmath23 sample indicate a mixture of two magnetic ordering at t@xmath37 @xmath4 70k and t@xmath38 @xmath4 225k , respectively , very similar to the cluster spin freezing temperature at t@xmath37 @xmath4 100k and short range ferrimagnetic ordering temperature at t@xmath38 @xmath4 230k , respectively , for bulk sample @xcite . the field dependence of magnetization data ( fig .
7 ) under zfc condition is similar to that of ferromagnetic isotherms but the magnetization lacks saturation even upto 12 tesla at any temperature .
this , we attribute to the spin canting effects at grain boundary @xcite or superparamagnetic contribution of nanoparticles @xcite .
we note from fig .
7a ( inset ) that the s@xmath20 sample has a better ferromagnetic behaviour in terms of saturation with respect to magnetic field which indicates that the strong spin canting behaviour of s@xmath19 sample @xcite has decreased for the nanoparticle s@xmath20 sample .
importantly , we also note ( fig.7a , inset ) that the saturation magnetization of nanoparticle sample s@xmath20 is significantly increased at room temperature with respect to the bulk sample s@xmath49 .
however , the higher value of m for s@xmath19 at 10k for h @xmath47 4 t compared to s@xmath20 can be understood by assuming two types of magnetic interactions in the system .
one is ferromagnetic and second one is antiferromagnetic . in s@xmath19 sample
the b site antiferromagnetic interactions are dominant which causes spin canting in b site and shows non - saturation in magnetization @xcite .
the reduction of antiferromagnetic interactions between the b site moments gives rise the better ferromagnetic behaviour , yet having a reduced saturation moment at 10k , for s@xmath20 sample .
it is observed from the hysteresis loop ( fig .
7b ) that isothermal remanent magnetization ( m@xmath50 ) , coercive field ( h@xmath51 ) where m@xmath50 is zero , and irreversible field ( h@xmath52 ) , where the hysteresis loop closes , decrease with temperature .
there is no hysteresis loop at t @xmath53 290k .
this suggest that the thermal energy close to the blocking temperature ( @xmath54 ) is sufficient to reduce the coercive field to zero value @xcite .
the saturation magnetization ( m@xmath55 ) values , obtained from m vs 1/h plot , are small in comparison with the bulk sample as seen in nanocrystalline nife@xmath2o@xmath3 @xcite .
this is possible if the b site intra - cluster spin canting decreases or if there is a redistribution of cations between a and b sites when the bulk sample is mechanically milled @xcite .
8 shows the room temperature mssbauer spectra in absence of any external magnetic field for the samples ( s@xmath19 ) , s@xmath20 , s@xmath21 , s@xmath22 and s@xmath23 .
the spectrum of s@xmath19 sample consists of a lorentzian doublet arising from the fe@xmath56 ions at octahedral ( b ) site and a lorentzian singlet arising from the fe@xmath56 ions at tetrahedral ( a ) site .
the most probable cations distribution of bulk sample ( s@xmath19 ) obtained is ( zn@xmath1fe@xmath0)@xmath12[co@xmath0fe@xmath57@xmath15o@xmath3 . the isomer shift ( is ) and quadrupole splitting ( qs ) of fe@xmath56 ions at [ b ] site are @xmath4 + 0.29 mm / sec and 0.35 mm / sec respectively .
the is value of fe@xmath56 ions at ( a ) site is @xmath4 -0.056 mm / sec @xcite .
the mssbauer spectrum of the bulk sample therefore confirm that zn@xmath58 ions occupy a site , co@xmath58 ions prefer b site and fe@xmath56 ions prefers both a and b sites .
the mssbauer spectrum of the s@xmath20 sample clearly indicates a hyperfine magnetic field splitting in addition to a central paramagnetic doublet .
this spectrum of s@xmath20 sample represents the appearance of spontaneous magnetization of ferromagnetic clusters mixed with superparamagnetic fluctuation effect due to nano meter size of the particles @xcite . as the annealing temperature increases , the hyperfine magnetic field splitting decreases for s@xmath21 sample and the s@xmath22 and s@xmath23 samples show paramagnetic spectra .
the mssbauer parameters of s@xmath23 sample ( is= + 0.27 mm / sec and qs = + 0.43 mm / sec for b site fe@xmath56 ions and the is = + 0.25 mm / sec of a site fe@xmath56 ions ) differ from that of the as prepared bulk sample which suggests that duration of heat treatment ( @xmath5 6 hours ) is not enough to achieve 100% equilibrium state of the bulk samples .
these results confirm that the enhancement of magnetic ordering and magnetization observed in s@xmath20 sample is intrinsic property of the material .
these also further indirectly confirm the hypothesis of the presence of non - equilibrium cation distribution in the as milled sample .
+ to check the recovery of cation distribution from non - equilibrium state to equilibrium state , we recorded the mssbauer spectrum of s@xmath20 sample as function of duration of heat treatment at 300@xmath6c .
9 shows the mssbauer spectra and corresponding hyperfine magnetic field ( hmf ) distribution ( p(h ) ) .
the presence of multiple hyperfine fields at a and b site fe@xmath56 ions can be expected due to various super transferred hyperfine fields ( sth ) from neighbouring ions .
the hmf acting on the b site fe@xmath56 ions is due to non - uniform environment , consisting of different number of zn@xmath58 and fe@xmath56 ions at the nearest neighbour a sites .
the hmf experienced by the a site fe@xmath56 ions is due to nearest neighbour b site configuration @xcite .
it is clear from the p(h ) distribution that the changes brought about by 2 hours heat treatment is not very significant but the peak intensities of hyperfine field components are decreasing and intensities of paramagnetic components are increasing for annealing time @xmath53 16 hours . indirectly , it gives the information that more and more number of zn@xmath58 ions stabilizing in a site with the compromise of more fe@xmath56 ions in b site .
this reduces the reduction of super transferred hyperfine magnetic field via inter - sublattice superexchange interactions @xcite .
10 shows the temporal evolution of the average hmf for 300@xmath6c heat treated sample .
the solid line represent the fit data using a function of time ( t ) as f(t ) = _ a _ +
_ b_(1-exp(t / t@xmath19 ) ) where _ a _ , _ b _ and t@xmath19 are constants .
f(t ) represents average hmf as a function of time .
the fit gives a measure of the reaction kinetics through atomic diffusion as the sample gets annealed .
usually nano materials lead to the decrease in magnetization and ordering temperature as the particle size decreases @xcite .
this is already observed in co@xmath0zn@xmath1fe@xmath2o@xmath3 with particle size @xmath5 6 nm to 70 nm , prepared by coprecipitation method @xcite .
however , the bulk co@xmath0zn@xmath1fe@xmath2o@xmath3 sample with short range ferrimagnetic ordering temperature @xmath5 260k and paramagnetic at 300k @xcite after 24 hours mechanical milling gives nanoparticles @xmath5 20 nm which shows ferromagnetic cluster blocking state below
t@xmath15 @xmath4 320k with an enhancement of magnetization . because of the weak inter - sublattice interactions j@xmath7 ( due to low a site fe@xmath56 population ) , the b site spins form canted structure in bulk co@xmath0zn@xmath1fe@xmath2o@xmath3 sample .
the stable cation distribution of the bulk sample is destroyed due to mechanical milling and a non - equilibrium disordered state , _
i.e. _ the increase of a site fe@xmath56 population enhances the inter - sublattice exchange interactions fe@xmath59-o@xmath9-fe@xmath60 ( j@xmath7 ) and change the b sublattice superexchange interactions fe@xmath60-o@xmath9-fe@xmath60 .
this makes the clusters more ferromagnetic with lower spin canting .
the reduction of spin canting inside the clusters enhances the magnetization .
the superparamagnetic fluctuation effect is expected due to the nano size clusters ( particles ) .
the cations redistribution surely is the main factor which controls the magnetic properties of nanoparticle system but distortion of spherical shape and strain induced anisotropy energy due to mechanical milling can give rise to a preferential orientation of the spins inside the clusters .
+ the magnetic reversibility of the heat treated ( s@xmath23 ) milled sample towards the bulk sample ( may not 100% , due to different micro structural parameters like particle size , cluster size distribution , spin canting angle inside the clusters ) conclusively show that enhancement of magnetization and ordering temperature of the as milled samples are intrinsic property of co@xmath0zn@xmath1fe@xmath2o@xmath3 spinel oxide .
+ acknowledgement : one of the authors rnb thanks council of scientific and industrial research ( csir , india ) for providing fellowship [ f.no.9/489(30)/98-emr-i ] .
d. j. fatemi , v. g. harris , m. x. chen , s. k. malik , w. b. yelon , g. j. long and a. mohan , j. appl .
phys . * 85 * , 5172(1999 ) j. p. chen , c. m. sorensen , k. j. klabunde , g. c. hadjipanayis , e. devlin and a. kostikas , phys .
b * 54 * , 9288(1996 ) g. f. goya and h. r. rechenberg , j. phys : condens .
matter * 10 * , 11829(1998 ) magnetic properties of fine particles , edited by j. l. dorman and d. fiorani ( north - holand , 1991 ) v. a. m. brabers , hand book of magnetic materials , edited by k. h. buschow * 8 * , 189(1995 ) ; s. krupika and p. novak , ferromagnetic materials , edited by e. p. wolfarth , north - holand publishing company * 3 * , 189(1982 ) m.h .
kryder , material res . bull .
, * 21 * , 17 ( 1996 ) scientific and clinical applications of magnetic carriers , edited by u. hfeli , w. schitt , j. tellet , m. zborowski , plenum : new york , ( 1997 ) s. k. khanna and s. linderoth .
* 67 * , 742(1991 ) p. crespo , a. hernando and a. garcia - escorial , phys .
b * 49 * , 13227(1994 ) h. h. hamdeh , j. c. ho , s. a. oliver , r. j. willey , g. oliveri and g. busca , j. appl . phys . *
81 * , 1851(1997 ) qi chen and z. j. zhang , appl .
. lett . * 73 * , 3156(1998 ) d. fiorani , j. l. dormann , r. cherkaoui , e. tronc , f. lucari , f. dorazio , l. spinu , m. nogues , a. garcia , a. m. testa , j. magn .
magn . mater .
* 196 - 197 * , 143(1999 ) l. bianco , a. hernando , e. bonetti and e. navarro , phys .
b * 56 * , 8894(1997 ) r. h. kodama , a. e. berkowitz , e. j. mcniff .
jr and s. foner , * 77 * , 394(1996 ) g. f. goya , h. r. rechenberg and j. z. jiang , j. magn .
magn . mater . * 218 * , 221(2000 ) m. zheng , x. c. wu , b. s. zou and y. j. wang , j. magn .
magn . mater . *
183 * , 152(1998 ) v. a. m. brabers , phys . rev .
lett . * 68 * , 3113(1992 ) p. j. van der zaag , a. noordermeer , m. t. johnson and p. f. bongers , phys .
* 68 * , 3112(1992 ) g. a. petitt and d. w. forester , phys .
b * 4 * , 3912(1971 ) r. n. bhowmik and r. ranganathan , j. magn . magn . mater .
( in press ) b. d. cillity , elements of x - ray diffraction ( second edition ) , addison - wesley publishing company , inc(1977 ) r. jenkins , r. gould and d. gedcke , quantitative x - ray spectroscopy ( dekker , new york ) , 445(1981 ) anindita ray , a. chakravarti and r. ranganathan , rev .
instrum . , * 67 * , 789 ( 1996 ) ; a. chakravarti , r. ranganathan and a. k. raychaudhuri , pramana-j.phys .
, * 36 * , 231 ( 1991 ) g. le car and j. m. dubois , j. phys .
e * 12 * , 1083(1979 ) j. a. mydosh , spin glasses : an experimental introduction ( taylor & francis ) , 1993 m. muroi , r. street , p. g. mccormick and j. amighian , phy .
b * 63 * , 184414 - 1(2001 ) a. e. berkowitz , j. a. lahut , c. e. vanburen , ieee transactions on magnetics * mag-16 * , 184(1980 ) c. n. chinnasamy , a. narayanasamy , n. ponpandian , k .
chattopadhyay , k. shinoda , b. jeyadevan , k. tohji , k .
nakatsuka , t. furubayashi and k. nakatani , phy .
b * 63 * , 184108 - 1(2001 ) * 12*,7795(2000 ) s. dey , a. roy , j. ghose , r. n. bhowmik and r. ranganathan , under communication fig.1 xrd spectra for co@xmath0zn@xmath1fe@xmath2o@xmath3 spinel oxide heated at different temperatures . for the notation s@xmath19 , s@xmath20 , s@xmath21 , s@xmath22 and s@xmath23 see in the text .
+ fig.2 x - ray fluorescence spectrum for the as milled sample .
the arrow indicates the corresponding metals characteristics k@xmath24 and k@xmath26 lines .
+ fig.3 ac susceptibity @xmath34 ( in a ) and @xmath35 ( in b ) data for as milled sample ( s@xmath20 ) .
t@xmath15 and t@xmath36 represent blocking and spin canting temperature , respectively .
+ fig.4 ac susceptibity @xmath34 ( in a ) and @xmath35 ( in b ) for the 1000@xmath18c-6 hour heat treated sample ( s@xmath23 ) .
fig.a inset shows ac susceptibility data at 1 oe ac field , f = 337 hz and ln f vs 1/t@xmath37 plot for the same sample .
+ fig.5 dc magnetization vs temperature for 24 hrs milled co@xmath0zn@xmath1fe@xmath2o@xmath3 sample .
b inset shows the temperature dependence of field cooled ( fc ) ( cooled in a field of 60 oe ) thermoremanent magnetization ( trm ) .
+ fig.6 temperature dependence of dc magnetization data for different heat treated samples at 100 oe field .
+ fig.7 a ) m vs h plot at different temperatures for s@xmath20 sample and inset fig .
compare the 10k and 290k data of s@xmath19 and s@xmath20 samples .
b ) hysteresis loopes at different temperatures for s@xmath20 sample . for m@xmath50 , h@xmath51 and h@xmath52
+ fig.8 room temperature mssbauer spectra recorded in absence of field for as prepared bulk ( s@xmath19 ) and heat treated as milled @xmath20 , s@xmath21 , s@xmath22 and s@xmath23 sample .
+ fig.9 room temperature mssbauer spectra for heat treated co@xmath0zn@xmath1fe@xmath2o@xmath3 sample at 300@xmath18c for different time ( left ) and corresponding hyperfine field distribution ( right ) .
+ fig.10 annealing time ( at 300@xmath18c ) dependence of average hyperfine field value for as milled co@xmath0zn@xmath1fe@xmath2o@xmath3 sample . | we report the magnetic properties of mechanically milled co@xmath0zn@xmath1fe@xmath2o@xmath3 spinel oxide . after 24 hours milling of the bulk sample , the xrd spectra show nanostructure with average particle size @xmath4 20 nm .
the as milled sample shows an enhancement in magnetization and ordering temperature compared to the bulk sample .
if the as milled sample is annealed at different temperatures for the same duration , recrystallization process occurs and approaches to the bulk structure on increasing the annealing temperatures .
the magnetization of the annealed samples first increases and then decreases . at higher annealing temperature
( @xmath5 1000@xmath6c ) the system shows two coexisting magnetic phases _
i.e. _ , spin glass state and ferrimagnetic state , similar to the as prepared bulk sample .
the room temperature mssbauer spectra of the as milled sample , annealed at 300@xmath6c for different durations
( upto 575 hours ) , suggest that the observed change in magnetic behaviour is strongly related with cations redistribution between tetrahedral ( a ) and octahedral ( o ) sites in the spinel structure . apart from the cation redistribution
, we suggest that the enhancement of magnetization and ordering temperature is related with the reduction of b site spin canting and increase of strain induced anisotropic energy during mechanical milling . |
quantifying the star formation histories of galaxies constitutes one of the major unsolved issues towards a complete understanding of galaxy formation .
although this task is difficult , analyses of stellar populations constitute a step forward toward the achievement of this goal . in the majority of cases
, these studies have to be made using integrated spectra or colours , as we can only resolve stars in a limited number of galaxies . while studies of stellar populations for early - type galaxies are abundant in the literature ( e.g. , trager et al .
2000 ; kuntschner 2000 , thomas et al . 2005 ; snchez - blzquez et al .
2006abc ; smith , lucy & hudson 2007 , among may others ) , these studies are much more sparse for disc galaxies . until very recently , the study of the stellar component in disk galaxies outside the local group was restricted to broadband photometric data ( bell & de jong 2000 ; macarthur et al .
2004 ; pohlen & trujillo 2006 ; muoz - mateos et al . 2007 ; roediger et al . 2012 ) .
these studies find that disc galaxies tend to be bluer in the external parts , a trend that is usually interpreted in terms of stellar population gradients as galaxies tend to be younger and more metal poor in the external parts .
however , there are large discrepancies in the magnitude of the stellar population gradients derived by different authors due , partially , to the difficulty in disentangling the effects of the age , metallicity and dust extinction by using only colours .
spectroscopic studies may help to alleviate the associated degeneracies but the low surface brightness of the disc region and the nebular emission lines filling some of the most important age - diagnostic absorption features make these studies very difficult . in the last few years , however , the advent of new techniques to separate the contributions of stellar and gaseous components ( see , e.g . , sarzi et al .
2005 ) and large aperture telescopes have allowed the obtention of enlightening results in this field ( see , e.g. , yoachim & dalcanton 2008 ; macarthur , gonzlez & courteau 2009 ; prez et al .
2011 ; snchez - blzquez et al .
2011 ) .
all the studies quoted above used long - slit spectroscopy , focusing along a predefined spatial axis .
disc galaxies , however , have a complex geometry , with many structural components , such as bulges , bars , rings , etc , which makes the interpretation of the long - slit data a difficult task .
integral field spectroscopy ( ifs , hereafter ) allows the study of the different galaxy components mentioned above separately .
ifs data cubes provide , for each pixel of the galaxy , a spectrum from which absorption and emission lines can be extracted , allowing to decipher the star formation history and the physical properties of the interstellar medium in galaxies .
also , ifs offers a unique opportunity to improve the signal to noise ratio of the data by averaging many more spectra in the external parts of the galaxy and makes easier the masking or the extraction of the hii regions in the spatial map . in the last few years
, there have been some studies of the stellar population properties of disc galaxies from ifs spectroscopy ( ganda et al . 2007 ; snchez et al .
2011 ; yoachim et al .
2009 , 2012 ) , probing the potential of this technique .
in fact , many ongoing and planned surveys are devoted to the observations of nearby galaxies using integral field units ( e.g. , califa ( snchez et al . 2011 ) ; venga ( blanc et al . 2010 ) ; sami ( croom et al . 2012 ) ; manga ( k. bundy ) ) .
a further complication in the study of stellar population in disc galaxies is that , surely , the stars did not form in a single burst of a given age and metallicity as usually assumed in studies of early - type galaxies .
therefore , the comparison of spectral characteristics with single stellar populations ( ssp ) models leads to information which is difficult to interpret .
fortunately , in the last decade there has been a tremendous improvement in the stellar population analysis techniques , associated with the release of well - calibrated stellar libraries ( e.g. , stelib , leborgne et al .
2003 ; miles , snchez - blzquez et al .
2006 ; indo - us ( valdes et al . 2004 ) ; cat ( cenarro et al .
2001ab ) that have allowed the construction of stellar population models that predict the whole synthetic spectrum of a population with a given age and metallicity ( e.g. , vazdekis 1999 ; vazdekis et al . 2003 , 2010 ; bruzual & charlot 2003 , conroy et al .
2009 , 2010ab ; maraston & strmbck 2011 ) .
this has driven the development of inversion algorithms to extract information about the star formation histories and the metallicity evolution from integrated spectra ( e.g. , moped : heavens et al .
2000 , panter et al . 2003 ; starlight : cid fernandes et al .
2005 , moultaka et al . 2005 ; steckmap : ocvirk et al . 2006ab
; vespa : tojeiro et al . 2007 ; fit3d : snchez et al .
( 2008 ) ; ulyss : koleva et al .
the combination of these two developments make the study of the stellar population properties in disc galaxies very timely . in the present paper
we derive the stellar population properties and star formation histories of ngc 628 in a 3-dimensional fashion , using several techniques .
ngc 628 ( m74 ) is an extensively studied isolated grand - design sa(s)c ( corwin et al .
1994 ) spiral galaxy at a distance of 9.3 mpc in the constellation of pisces .
it does not show any evidence of having had encounters with satellites or other galaxies in the last @xmath1 yr ( kamphuis & briggs 1992 ) .
regarding its structure , ngc 628 shows an outer hi ring at around 12 arcmin from the nucleus ( roberts 1962 ; briggs et al .
1980 ) that seems to be due to the interaction with two large high - velocity clouds accreting onto the outer parts of the disc ( kamphuis & briggs 1992 ; lpez - corredoira et al . 2002
; beckman et al .
it also has a inner rapidly rotating disc - like structure ( baigle et al .
2006 ) , a circumnuclear ring of star formation at @xmath02 kpc from the center ( discovered in @xmath2co j=1 - 0 sub - mm , wakker & adler 1995 , and 2.3 @xmath3 m co absorption , james & seigar 1999 ) and a nuclear , nested bar on a @xmath0100 pc scale ( laine et al .
an oval structure has also being detected at a radii around 50@xmath4 kpc ( seigar 2002 ) .
all these characteristics indicate that secular evolution processes are taking place in this galaxy ( see fathi et al .
2007 ) .
this paper can be considered as a pilot study that allows us to explore the techniques that will be applied to the study of a larger sample of galaxies from the califa survey ( snchez et al .
the study complements and expands the ones presented in snchez et al .
( 2011 , paper i ) and rosales - ortega et al .
( 2011 , paper ii ) for the same galaxy .
ngc 628 is the largest object within the ppak integral field spectroscopy nearby galaxies survey : pings ( rosales - ortega et al.2010 ) .
pings ifs survey of nearby ( @xmath5100 mpc ) well - resolved spiral galaxies is specially designed to obtain complete maps of the emission - line abundances , stellar populations , and reddening using an ifs mosaicking imaging , which takes advantage of what is currently one of the world@xmath6s widest field - of - view ( fov ) integral field units .
section [ sec_obs ] briefly describes the observations , sec .
[ sec : analysis ] the analysis performed to derive the stellar population properties and , in sec . [ sec_results ] we present our results .
section [ sec : comparison ] tests the robustness of the solution comparing with the analyses performed with different techniques and by different authors . in sec .
[ sec_discussion ] we provide a brief discussion and sec .
[ sec_conclussion ] summarizes our conclusions .
the observations analysed here are part of the ppak ifs nearby galaxies survey ( pings , rosales - ortega et al .
the ppak fibre bundle consists of 382 fibers of 2.7@xmath7 diameter each . of these 382 fibers ,
331 are concentrated in a single hexagonal bundle covering a field - of - view of 72@xmath863@xmath7 , with a filling factor of @xmath065% .
the sky background is sampled by 36 additional fibers in 6 bundles of 6 fibers each , distributed along a circle of @xmath090 arcsec from the center . due to the large size of ngc 628 compared to the field - of - view of the instrument ,
a mosaicking scheme was adopted .
the central position was observed in dithering mode to gain spatial resolution , while the remaining positions were observed without dithering due to the large size of the mosaic .
the observations for this galaxy extended over a period of three years in different stages , with a total of six observing nights .
the spectroscopic mosaic contains 11094 spectra covering an area of @xmath034 arcmin@xmath9 ( see papers i and ii for details ) , which is the largest area ever covered by an ifu mosaicking .
the v300 grating was used for all the observations , covering the wavelength range @xmath03700 - 7100 , with a spectral resolution of fwhm@xmath08 .
the seeing varied between @xmath01 and @xmath01.8 arcsec , and the median seeing was 1.4 arcsec .
different spectrophotometric stars were observed during the observing runs with , at least , two stars observed each night .
the estimated spectrophotometric accuracy is @xmath00.2 mag .
figure [ mosaico_n628 ] shows the mosaic pattern covered by the observations .
the data reduction is described in rosales - ortega et al .
( 2010 ) and will not be repeated here .
the main properties of the galaxy are shown in table [ tab : n628 ] . [ cols="<,^,^,^,^,^,^,^,^ " , ] one of the most important problems affecting stellar population studies is the well known age - metallicity degeneracy , i.e .. , the colours and spectral characteristics of a given population can be mimicked with another population younger but more metal rich or older but more metal poor . to explore upon what extent
we are affected by this degeneracy , we use the integrated spectrum of ngc 628 and added noise to simulate spectra of signal - to - noise 20 , 40 and 70 ( 70 is the s / n of the original spectrum ) .
we performed 2500 montecarlo simulations in which each pixel was perturbed with the associated noise error , following a gaussian distribution . for each simulation
a new value of the age and metallicity was derived .
the results are shown in fig .
[ chi_map ] .
the mean values obtained and the root - mean square dispersion ( rms ) are shown in table [ table : simus ] . as can be seen , as the noise increase , also the uncertainty in the parameters and , therefore , the age - metallicity degeneracy .
furthermore , there seem to be a small systematic effect in the sense that as the noise increase , the mean derived age decrease .
we decided to bin our data with a minimum signal - to - noise of 40 ( see sec .
[ sec_maps ] ) as a compromise between obtaining good quality spectra and resolve spatially the different components of the galaxy .
the typical dispersion in the mean luminosity - weighted age and metallicity derived in a spectrum with this signal - to - noise is 0.05 , and slightly larger for the mass - weighted means ( 0.08 and 0.06 for the age and metallicity values respectively ) .
figure [ fig_sfr ] shows the typical steckmap outputs for the integrated spectra of ngc 628 .
the flux and mass fraction and the derived age metallicity relation shows that the majority of the stars in this galaxy formed @xmath1013 gyr ago and less than 10% of stars did it @xmath0 1 gyr ago .
note , that the maximum age of the models used here is 17.78 gyr , ie .
, exceeding the age of the universe in the adopted cosmology .
it goes beyond the scope of this paper discussing the reasons for these discrepancies and we just caution the reader that the absolute values of the old populations have to be taken as approximate values . in any case
, this does not affect any of our conclusions as we will only discuss relative stellar population variations along the galactic radius and relative trends between very old and young populations .
in order to reach the desired signal - to - noise ratio for our analysis to ensure an appropriate modeling of the stellar populations we used the voronoi tessellation algorithm of cappellari & copin ( 2003 ) and the pingsoft ( rosales - ortega 2011 ) software to bin the data cube in regions with a minimum signal - to - noise ( s / n ) ratio of 40 ( see sec . [
sec : steckmap ] ) , over a spectral region centered at 5400 with a 100 width . figure [ voronoi_binning ] shows the different regions binned in our scheme of the voronoi - tessellation algorithm .
we run steckmap on each of the binned spectra and obtained star formation histories and age - metallicity relations for all of them .
once we have derived star formation histories , we can obtain average values weighted by the flux or by the mass of the population . while the mean values weighted with mass are going to be biased towards the parameters of the majority of the stars , the flux - weighed values are biased to the parameters of the youngest component , as young stars are much more luminous in the wavelength range we are studying .
figure [ fig_maps ] shows the luminosity- and mass - weighted maps of age and metallicity for ngc 628 .
several things can be appreciated in these maps .
the mass - weighted mean age map shows that the majority of the stars in the galaxy are old , even in the most external regions sampled by the data , while the luminosity weighted age map reveals younger population in the external parts . in the mass - weighted age map a ring of younger ( with an average age of @xmath01.7 gyr ) population
is also clearly visible with yellowish colors , coincident with a young ring that has also been detected in co ( wakker & adler 1995 ) .
the mass - weighted metallicity map shows that the ring and the central parts of the galaxy are more metal poor than the external parts , which is compatible with the idea that the central ring has been formed as a consequence of the inflows of gas from the external parts ( e.g. , kormendy & kennicutt 2004 ; snchez et al .
2011 ) . in the next section ,
we quantify these trends . to obtain radial trends of the stellar population parameters , we average the individual spectra in concentric annuli , which allows to obtain the required signal - to - noise ratio for our analysis up to @xmath0 120 arcsec ( @xmath00.4@xmath11 or @xmath05.4 kpc ) in linear projected galactocentric radii .
we obtained azimuthally averaged spectra by co - adding all the spectra within successive rings of 4 arcsec .
these azimuthally averaged spectra were then analysed using the fitting procedure described above .
following zou et al .
( 2011 ) we will consider the disc the region r@xmath1230
arcsec , while the bulge that of [email protected] arcsec .
figure [ fig_sfr1 ] shows the mean flux fraction as a function of age and the age - metallicity relation calculated at different distances from the center of the galaxy .
it can be seen that , in the center , there is a younger population ( @xmath02 gyr ) that dominates the flux while its contribution gets diluted as we go to larger radii .
we also see an increase in the metallicity with time , as it will be expected in a normal chemical evolution model .
+ + + + to quantify these trends we show , in figure [ gradients ] , the mass and luminosity - weighted age and metallicity gradients for ngc 628 . in both
, age and metallicity profiles , there seems to be change of slope at around @xmath060 and @xmath080 arcsec respectively .
we will analyse the gradients inside and outside this region separately .
the mass - weighted age of this galaxy is old ( @xmath010 gyr ) at all sampled radii , i.e. , the disc is dominated in mass by old stars , even at @xmath02 scale - length .
this is in agreement with snchez - blzquez et al .
( 2011 ) , where it was found that the mean , mass - weighted age of 4 disc galaxies was also around @xmath0 10 gyr , at radii larger than 2 scale - length of the discs ( see also macarthur et al .
2009 ) .
the luminosity - weighted mean age is biased towards the age of the youngest components , as these components are much more luminous in the wavelength range we are studying . here , it can be seen , as we have already mentioned , that the age gradient show two different behavior . in the inner disc the gradient is flatter ( or even positive , with age increasing with radius ) .
a minimum in the mean luminosity - weighted age can also be seen in at @xmath010@xmath7 ( this is more readily seen in the azimuthally weighted values , where the errors are smaller ) , coincident with the radius of a circumnuclear star forming ring previously detected in sub - mm co ( 1 - 0 ) observations ( wakker & adler 1995 ) and infrared 2.3 @xmath3 m co absorption ( james & seigar 1999 ) and was also seen by ganda et al .
( 2006 ) in their h@xmath13 and [ oiii ] maps of sauron . in the outer disc
, the luminosity - weighted age shows a clear gradient , with a variation of almost @xmath06 gyr in 2 scale - lengths .
previous studies of the stellar population across the radius for this galaxy include those of natali et al.(1992 ) and zou et al .
( 2011 ) using photometry in different bands and that of snchez et al.(2011 ) , using spectroscopy .
zou et al .
( 2011 ) found , as in the present study , a different two distinct disc components , on from 30 to 60 arcsec , and the second one from 60 to 132 arcsec . at 1@xmath6 , the age profile change slope .
we also find a change of slope at a similar radius but the trends that we derive are very different in both studies .
zou et al .
( 2011 ) found that the inner region of the disc ( from 30 - 60@xmath7 ) have a much steeper gradient than the outer region , which is completely the opposite to what we find in the present study .
our profile , however , coincides very well with that of snchez et al .
( 2011 ) ( see fig .
[ fig : sanchez ] ) .
the metallicity gradient also shows two differentiated regions , being almost flat or slightly positive in the central parts and negative outwards .
the radius at which the metallicity gradient changes slope is similar to the break calculated by scarano & lpine ( 2013 ) for the gas - phase metallicity using data from rosales - ortega et al .
( 2011 ) , and also similar to the break in the stellar metallicity found by zou et al .
in fact , qualitatively , the trends found by zou et al .
( 2011 ) are very similar to our ; in the disc region the metallicity increase with radius until @xmath14 and decrease slightly beyond .
our gradient in the internal regions is , however , much milder and the overall metallicities considerably larger those that derived by the quoted authors and the radius at which the gradient change slope is larger .
this dual behavior is , however , not reported by snchez et al .
( 2011 ) , although their luminosity - weighted values of age and metallicity are in a very good agreement with ours ( see fig . [
fig : sanchez ] ) . to check the reliability of this break we compare the metallicities obtained with our methods with those obtained with a better understood and widely used method , a classical index - index diagram .
figure [ compara ] shows the comparison of the metallicity gradient obtained with steckmap and the one obtained using an index - index diagram combining h@xmath13 and [ mgbfe ] .
we are not comparing exactly the same quantity as the metallicity obtained by using an index - index diagram is a single - stellar population equivalent value , while the one obtained with steckmap is a luminosity - weighted one .
however , it has been shown that these two measurements give very similar values ( serra & trager 2007 ) .
it can be seen that , although the ssp - equivalent values are noisier , the trends obtained with the two methods are very similar and , in particular , a change of slope is visible at the same radius . the different trends in age and metallicity between this study and that of zou et al . ( 2011 ) shows the difficultly in deriving accurate stellar population parameters using broad band colors , due to the existent degeneracies between age , metallicity and dust extinction .
it is worth noticing that the broad band optical color profiles of this galaxy show , in fact , a very steep slope in the inner region of the disc while it flattens out beyond 1@xmath6 .
on the contrary , we obtain much flatter colors of age and metallicity in the inner regions of the disc .
we prove , however , in sec .
[ comp : photometry ] that we can reproduce the observed colors of this galaxy using our derived star formation histories .
one of the advantages of being able to derive an age - metallicity relation is that it allows us , in principle , to obtain the time evolution of the metallicity gradient .
the temporal evolution of the metallicity gradient is an issue that is not yet settled from neither , the theoretical or the observational point of view .
some chemical evolution models predict a steepening with time starting from initially inverted or flat gradients ( e.g. chiappini , matteucci & romano 2001 ) while others predict an initially negative gradient that flattens ( e.g. , moll & daz 2005 ) .
recent work ( pilkington et al . 2012 ; gibson et al .
2013 ) have showed that , in cosmological simulations , the evolution of the metallicity gradient with time depends on the details of the sub - grid physics implemented in the hydrodynamical codes of galaxy formation , in particular , on the feedback scheme . therefore , the study of the metallicity gradient for different age population in galaxies are key to reveal the formation processes of disc galaxies and better constrain the physics included in numerical simulations .
some authors have studied this evolution in our own mw using planetary nebulae ( e.g. , maciel et al .
2003 ; stanghellini & haywood 2010 ; maciel & costa 2013 ) also with discrepant results .
a direct measurement of the evolution of the metallicity gradients can be obtained analysing the magnitude of the gas - phase metallicity gradient at high redshift ( cresci et al .
2010 ; jones et al . 2010 ; yuan et al .
2011 ; queyrel et al .
yuan et al .
( 2011 ) found that , for at least one grand design spiral at redshift z 1.5 , the metallicity gradient is significantly steeper ( @xmath15 dex / kpc ) than the typical gradient encountered today .
however , the difficulties in obtaining high resolution data for likely mw progenitors has mean that the theoreticians have had very few constraints on their models , as different paths can lead to a similar final metallicity gradient .
figure [ evol_grad ] shows the spatially resolved metallicity in the hii regions of this galaxy ( taken from rosales - ortega et al .
2011 ) and the stellar components younger than 2 gyr and older than 5 gyr . to compare the gas- and stellar metallicities we have transformed the former to the scale of the later assuming a solar metallicity of 12+log(o / h)=8.7 ( asplund et al .
it can be seen that , if we restrict ourselves to the younger stars ( those with ages@xmath52 gyr ) , the gradient shows a linear behavior , decreasing slightly with radius , and it is very similar to the gradient of the ionized gas . on the contrary , the metallicity profile for the older stars
have a different slope inside and outside @xmath050 - 60@xmath16 .
the metallicity gradient for the older populations is slightly positive in the internal zones , but it is steeper than that of the young population in the external parts . note that , although in paper i and in scarano & lpine ( 2013 ) a change of slope in the gas - phase metallicity at about the same radius as the one found here for the stellar component is reported , this is much more subtle than the one observed in the metallicity gradient of the old stars ( fig . [ evol_grad ] ) .
snchez et al . ( 2011 ) presented an analysis of the stellar population properties for the same galaxy using also the same data as the one presented here .
these authors used fit3d ( snchez et al .
fit3d enables linear fits of a combination of single stellar populations , and non - linear ones of emission - lines plus an underlying stellar population .
it also includes the cardelli , clayton & mathis ( 1989 ) attenuation law ( see snchez et al .
2011 for details ) . in snchez et al . ( 2011 ) ,
fit3d was used in combination with bruzual & charlot ( 2003 ) models , assuming a salpeter imf .
the comparison between their results and those of the present study are shown in fig .
[ fig : sanchez ] .
it is remarkable the excellent agreement between both studies , despite the differences in the code and in the stellar population models used .
this agreement gives particular confidence in the results .
ngc 628 has been extensively studied by many authors .
we can take advantage of the available data to check the methodology presented in this paper .
natali et al .
( 1992 ) performed surface photometry of this galaxy and presented u - v and b - v colors along the radius corrected by extinction .
one way to test the viability of our employed inversion technique is to use our derived star formation histories and age - metallicity relation to infer a colour gradient , by adding the colour corresponding to each population weighting with the luminosity in the corresponding photometric band .
we did this , and the results are shown in fig .
[ color_predicted ] .
as can be seen , the agreement is very good in the radial region covered by our data .
it is clear from previous studies that ngc 628 is a perfect candidate for a galaxy having experienced secular evolution . its bulge has the characteristic of a bulge formed secularly ( kormendy & kennicutt 2004 ) , with a surface brightness profile showing a srsic index @xmath52 ( zou et al .
2011 ) , presence of nuclear spiral arms ( cornett et al . 1994 ) and a young ring structure , a nuclear velocity dispersion drop ( ganda et al .
fathi et al . ( 2007 ) found indications that the gas is falling in from the outer parts towards the central regions , where a nuclear ring has formed at the location of the inner lindblad resonance radius of an m=2 perturbation .
we have analysed the stellar population gradients in the galaxy ngc 628 and found that the stellar population gradients in the disc shows a two distinct regions , one older and more metal rich with almost flat gradient in metallicity and slightly positive in the mean ( luminosity - weighted ) age and an external part with slightly decreasing gradients in both , luminosity - weighted age and metallicity .
the mass - weighted age gradient is compatible with being flat in all the regions sampled by our data , indicating that this galaxy is dominated , in mass , by old stars ( @xmath129 gyr ) .
outside r@xmath01@xmath6 , the analysis above shows that the majority of the stars in the disc of this galaxy are old , even at @xmath0 2 scale - lengths of the disc , but that the proportion of young stars is higher at larger radii , i.e. , the outer parts of the disk formed a larger fraction of their stars at recent times than the inner parts , consistent with the @xmath7inside - out@xmath7 growth scenario .
however , even the outer regions sampled by us , contain a large percentage of old stars , reflected in an old mass - weighted mean age .
studies of stellar populations using resolved stars have found old stellar populations in the outskirts of disc galaxies ( e.g. , ferguson & johnson 2001 ( m31 ) ; davidge 2003 ( ngc2403 , m33 ) ; galleti , bellazzini & ferraro 2004 ( m33 ) ; gogarten et al .
( 2010 ) ( ngc300 ) ) .
it has been suggested that these properties are not expected in cdm models ( ferguson & johnson 2001 ) where the disc grows inside - out and at relatively recent epochs ( @xmath17 1 ) .
however , in snchez - blzquez et al .
( 2009 ) , we analysed a fully cosmological , hydrodynamical disc galaxy simulation in a @xmath18cdm universe finding , also , a large percentage of old stars in the outskirts .
on the other hand , numerical simulations have shown that radial migrations due to the presence of non - axisymmetric components , like bars or spiral arms , can make that an important percentage of stars currently in the outer parts of the disc may have formed closer to the center of the disc and migrated outward to their current locations .
therefore , an observed old stellar population in the outer disc may not necessarily indicate that the outer disc formed early . in the present study , we have been able to analyse , for the first time , the evolution of the stellar metallicity gradient with time from integrated spectra . in the @xmath7outer@xmath7-disc region
, we have found that the metallicity gradient for the old stellar population is steeper than that of the young component .
this , in principle , can be a sign that stellar migration has not been very important in the disc region of this galaxy . however , to conclude this we would need to know the original _ gas - phase _ metallicity gradient at the epoch of formation of these old stars , which could have been even steeper . at radius @xmath19 both , the mean age and metallicity profiles , are flatter than in the external disc .
in fact , the gradient is even positive for the luminosity - weighted values of age and metallicities .
this is difficult to explain in an inside - out formation scenario for the disc formation and requires an explanation .
the change in the slope of the gas - phase metallicity gradient has been observed before in many galaxies ( e.g. , zaritsky 1992 ; vila - costas & edmunds 1992 ; martin & roy 1995 ; dutil & roy 1999 ; roy & walsh 1997 ; zahid & bresolin 2011 ; scarano & lpine 2013 ) and it is usually interpreted as the result of variations of gas density , the large - scale mixing induced by a bar ( friedli & benz 1995 ) , or by the spiral arms . in this case , the change of slope coincide with the corotation radius of the spiral patterns ( e.g. , scarano , lpine & marcon - uchida 2011 ; scarano & lpine 2013 ) .
this is because the main gas flows have opposite directions inside and outside corotations .
however , the above mechanism applies to young objects , such hii regions , that reveal the present value of the metallicity in the interstellar medium the fact that the break in the stellar metallicitiy is more pronounced in the older component points toward dynamical effects ( radial migrations ) as the cause for the observed dual behavior .
flat gradients , as the ones found here in the inner disc require strong radial mixing , which is difficult to produce with star formation histories but can be achieved in ongoing strong galaxy interactions ( rupke et al .
2010 ) or can be produced by secular evolution in the presence of non - axisymmetric components , like bars , oval or spiral arms .
ngc 628 is classified as an unbarred galaxy and , therefore , if secular evolution has happened , one is tempted to attribute it to the presence of spiral arms .
strong resonant interaction with transient or long - lived spiral arms at the corotation radius can produce stellar wanderings of several kpc in a few hundred myr ( e.g. , sellwood & binney 2002 ; lpine , acharova & mishurov 2003 ; rokar et al .
if the spirals are long - lived , this mechanism has been shown to produce bimodal patterns in the metallicity gradients with breaks at the spiral corotation radius .
the corotation radius for ngc 628 is located around @xmath07 kpc ( sakhibov & smirnov 2004 ) which is outside the region sampled by us .
therefore , the break that we observe here is too internal to be produced by the spiral arms .
however , a change on the slope of the color gradients around this radius has been reported in natali et al .
( 1992 ) . on the other hand , despite ngc 628
is classified as an unbarred galaxy in de vaucouleurs et al .
1991 atlas ) , near - infrared imaging and isophotal analysis has revealed the existence of an oval distortion with radius of r@xmath050@xmath16 ( laine et al . 2002 ; seigar 2002 ) confirmed by the kinematical analysis by fathi et al .
these authors also conclude that this oval ( a m=2 perturbation ) is the most prominent in the observed velocity field .
thought ovals are rather frequent in disk galaxies , their exact nature has not yet been investigated .
kormendy ( 1979 ) hypothesized that they are products of bars decays , which is supported by some observations of galaxies with prominent rings ( e.g. silchenko & afanasiev 2002 ) .
such a scenario is also predicted by the models of friedli & benz ( 1993 , 1995 ) ; combes et al .
1990 ; raha et al .
1991 ; martinez - valpuesta & shlosman 2004 .
the presence of a circumnuclear ring of star formation ( wakker & adler 1995 ; james & seigar 1999 ) and a central @xmath20-drop ( ganda et al . 2006 ) in ngc 628 support this hypothesis , as the large - scale oval perturbation can transport material from the outer zones to regions near the center of the galaxies ( see fathi et al .
2007 for discussion ) .
our results also show that we have found that the stellar population in the region dominated by the oval distortion reveals a positive luminosity - weighted age and metallicity gradients .
this has been observed in bars ( prez , snchez - blzquez & zurita 2009 ) and , in fact , moorthey & holtzman ( 2006 ) , found that the only galaxies with positive metallicity gradients in the central regions were those hosting bars .
positive stellar population gradients have also been reproduced in chemodynamical simulations of barred galaxies ( wozniak et al .
2007 ) .
weak oval structures are not believed to have a strong influence in the redistribution of stars inside the galaxy . however , this structures could have been stronger in the past . in fact , some authors ( kormendy 1979 ; 1981 , 1982 ; combes 1996 ) suggest that some bars evolve into lens components . therefore , the arguments above suggest that we could be witnessing a bar in the process of being destroyed , support coming from previous kinematical studies and from our stellar population analysis .
understanding bar formation and evolution is crucial to understand galaxy evolution , yet a fundamental question , namely whether bars can be destroyed by internal processes , remains unanswered .
from the theoretical point of view , there is currently a debate in the literature . on one hand ,
bars seem to be robust structures which can not be easily destroyed by central mass concentration as the observed are not massive enough ( shen & sellwood 2004 ; athanassoula , lambert & dehnen 2005 ) . on the other hand
some numerical simulations predict that bars can be destroyed and reformed when disc are very gaseous , and that galaxies can have up to 3 - 4 generations of bars over a hubble time ( bournaud & combes 2002 ) .
recent observations are consistent with the picture in which bars in early - type disc galaxies are long - lived , whereas those in late - type spirals could be dynamically young ( gadotti & de souza 2005 , 2006 ; elmegreen et al . 2007 ; sheth et al . 2008 ) .
however , there is no observational result to date that indicates whether bars can dissolve . in particular
, we can not tell whether an unbarred galaxy has had a bar in the past .
however , ovals ( or lenses ) can give us some clues , as one of the proposed formation scenarios is that lenses could be the remnants of a dissolved bar .
here we show stellar population gradients that are compatible with this scenario as , otherwise , it would be difficult to obtain the change on the slope of the stellar population parameters at the radius we are observing them and the slightly positive metallicity gradient in the inner disc .
in this paper we have presented an analysis of the star formation history and metallicity evolution in a 2-dimensional fashion for the galaxy ngc 628 , using data form the survey pings ( rosales - ortega et al .
2010 ) with the main purpose of validate our techniques to derive reliable stellar population properties .
* we derive 2d - star formation histories for ngc 628 using the code steckmap .
the analysis allows to obtain mean ages and metallicities weighted with both , luminosity and stellar mass .
* we check the robustness of the derived mean ages and metallicities to different choices of the fit . in particular
, we have checked the effect of deriving simultaneously kinematics and metallicities , and also the effect or fitting or not the stellar continuum .
we have found that , when the kinematics is fitted simultaneously to the derivation of star formation histories , the ages are not very affected , but the derived metallicities are . we have also found that the differences can be minimized in spectra with enough signal - to - noise . *
we find a negative luminosity - age gradient compatible with an inside - out formation scenario ( e.g. , larson 1976 ; white & frenk 1991 ; mo et al .
1998 ; naab & ostriker 2006 ; brook et al .
however , even at @xmath02.0 scale - lengths the disc ( assuming a disc scale - length of 3.2 kpc
zou et al .
2011 ) is dominated , in mass , by old stars ( @xmath010 gyr ) .
* we find a metallicity gradient that is flatter in the internal region and change to a negative slope in the external parts .
the break in the metallicity gradient is much more prominent in the old stars than in the younger components , which gradient is very similar to the gas - phase metallicity gradient .
however , a break is visible in all the different components at about the same radius .
we speculate that the metallicity gradient is the consequence of the gaseous flows and stellar migration produced by a @xmath21 perturbation , either the spiral pattern of a bar .
however , the position of the break is more coincident with the corotation radius of the oval distortion than that of the spiral pattern , which is beyond the radius sampled by our data .
as the oval distortion is ngc 628 is too weak for this to happen very efficiently , it could be that there was a stronger bar present in this galaxy and that the oval perturbation is the consequence of the dissolution of this structure .
we also argue that this structure can not have evolved much in size in , at least , the last @xmath0 5 gyr .
this work has received support form the spanish ministerio de ciencia e innovacin under the research projects aya2010 - 21322-c03 - 03 ( ravet ) and aya2010 - 21887-c04 - 03 ( estallidos ) and aya2012 - 31935 .
psb and sfs acknowledge support from the ramn y cajal program of the spanish ministerio de economa y competitividad .
ffro acknowledges the mexican national council for science and technology ( conacyt ) for financial support under the programme estancias posdoctorales y sabticas al extranjero para la consolidacin de grupos de investigacin , 2010 - 2011 .
based on observations collected at the centro astronmico hispano - alen ( caha ) at calar alto , operated jointly by the max - planck institut fr astronomie and the instituto de astrofsica de andaluca ( csic ) .
99 , m. , grevesse , n. , sauval , a. j. , allende prieto , c. , kiselman , d. , 2005 , a&a , 435 , 339 athanassoula , e. , lambert , j. c. , dehnen , w. , 2005 , mnras , 363 , 496 , j. e. , lpez - corredoira , m. n. , betancort - rijo , j. , castro - rodrguez , n. , cardwell , a. , 2003 , astronomical society series , 284 , 747 , e. f. , de jong , r. s. , 2000 , mnras , 312 , 497 bender r. , 1990 , a&a , 229 , 441 blanc et al .
, 2010 , in astronomical society of the pacific conference series , vol .
432 , new horizons in astronomy : frank n. bash symposium 2009 , stanford l.m . ,
green j.d .
, hao l. , mao y. , eds .
, p.180 boroson t. , 1981 , apjs , 46 , 177 bournaud f. , combes f. , 2002 , a&a , 392 , 83 briggs , f. h. , wolfe , a. m. , krumm , n. , salpeter , e. e. , 1980 , 238 , 510 , c. b. , kawata , d. , martel , h. , gibson , b. k. , bailin , j. , 2006 , apj , 639 , 126 , g. , charlot , s. , 1993 , apj , 405 , 538 , g. , charlot , s. , 2003 , mnras , 344 , 1000 , m. , copin , y. , 2003 , mnras , 342 , 345 , m. , emsellem , e. , 2004 , pasp , 116 , 138 , j. a. , clayton , g. c. , mathis , j. s. , 1989 , apj , 345 , 245 , a. j. , cardiel , n. , gorgas , j. , peletier , r. f. , vazdekis , a. , prada , f. , 2001 , mnras , 326 , 959 , a. j. , gorgas , j. , cardiel , n. , pedraz , s. , peletier , r. f. , vazdekis , a. , 2001b , mnras , 326 , 981 , a. j. , peletier , r. f. , snchez - blzquez , p. , selam , s. o. , toloba , e. , cardiel , n. , falcn - barroso , j. , gorgas , j. , jimnez - vicente , j. , vazdekis , a. , 2007 , mnras , 374 , 664 , c. , matteucci , f. , romano , d. , 2001 , apj , 554 , 1044 , i. v. , prugniel , p. , silchenko , o. k. , afanasiev , v. l. , 2007 , mnras , 376 , 1033 , r. , mateus , a. , sodr , l. , stasiska , g. , gomes , j. m. , 2005 , mnras , 358 , 363 mateus , a. and torres - papaqui , j. p. and schoenell , w. , 2007 , mnras , 375 , l16 , f. , debbasch , f. , friedli , d. , pfenniger , d. , 1990 , a&a , 233 , 82 combes f. , in astronomical society of the pacific conference series , vol .
91 , iau colloq . 157 : barred galaxies , buta r. , crocker d. a. , elmegreen b. g. , eds . ,
p. 286 , f. , debbasch , f. , friedli , d. , pfenniger , d. , 1990 , a&a , 233 , 82 , c. , gunn , j. e. , white , m. , 2009 , apj , 699 , 486 , c. , white , m. , gunn , j. e. , 2010 , 708 , 58 , c. , gunn , j. e. , 2010 , apj , 712 , 833 , r. h. , oconnell , r. w. , greason , m. r. , offenberg , j. d. , angione , r. j. , bohlin , r. c. , cheng , k. p. , roberts , m. s. , smith , a. m. , smith , e. p. , talbert , f. d. , stecher , t. p. , 1994 , apj , 426 , 553 , g. , mannucci , f. , maiolino , r. , marconi , a. , gnerucci , a. , magrini , l. , 2010 , nature , 467 , 811 , 2012 , astro - ph 1122.3367 , t. j. , 2003 , aj , 125 , 3046 , y. and roy , j .-
, 1999 , apj , 516 , 62 , y. and beauchamp , d. and roy , j .-
, 1995 , apjl , 444 , l85 , b. g. , 2007 , bars , spiral structure , and secular evolution in disk galaxies , de jong r.s . , ed . , p.163 , k. , beckman , j. e. , zurita , a. , relao , m. , knapen , j. h. , daigle , o. , hernandez , o. , carignan , c. , 2007 , a&a , 466 , 905 , a. m. n. and johnson , r. a. , 2001 , apj , 559 , friedli , d. and benz , w , 1993 , a&a , 268 , 65 friedli d , benz w. , 1995 , a&a , 301 , 649 gadotti d. a. , de souza r. e. , 2005 , apj , 629 , 797 gadotti , d. a. , de souza r. e. , 2006 , apjs , 163 , 270 galleti , s. , bellazzini , m. , & ferraro , f. r. , 2004 , a&a , 423 , 925 , k. and falcn - barroso , j. , peletier , r. f. , cappellari , m. , emsellem , e. , mcdermid , r. m. , de zeeuw , p. t. , carollo , c. m. , 2006 , mnras , 367 , 46 , k. , peletier , r. f. , mcdermid , r. m. , falcn - barroso , j. , de zeeuw , p. t. , bacon , r. , cappellari , m. , davies , r. l. , emsellem , e. , krajnovi , d. , kuntschner , h. , sarzi , m. & van de ven , g. , 2007 , mnras , 380 , 506 , b. k. , pilkington , k. , brook , c. b. , stinson , g. s. , bailin , j. , 2013 , a&a , 554 , 47 gil de paz , a. , madore , b. f. , boissier , s. , thilker , d. , bianchi , l. , snchez contreras , c. , barlow , t. a. , conrow , t. , forster , k. , friedman , p. g. , martin , d. c. , morrissey , p. , neff , s. g. , rich , r. m. , schiminovich , d. , seibert , m. , small , t. , donas , j. , heckman , t. m. , lee , y .- w . , milliard , b. , szalay , a. s. , wyder , t. k. and yi , s. gogarten , s. m. , dalcanton , j. j. , williams , b. f. , rokar , r. , holtzman , j. , seth , a. c. , dolphin , a. , weisz , d. , cole , a. , debattista , v. p. , gilbert , k. m. , olsen , k. , skillman , e. , de jong , r. s. , karachentsev , i. d. , quinn , t. r. heavens a. f. , jimenez r. , lahav , o. , 2000 , mnras , 317 , 965 , e. b.,lauberts , a.,schuster , h .- e . , west , r. m. , 1975 , a&as , 1975 , 22 , 327 , c. g. , walterbos , r. a. m. , bothun , g. d. , 2001 , apj , 559 , 878 huang , m .-
kauffmann , g. , chen , y .- m . ,
moran , s. m. , heckman , t. m. , dav , r. & johansson , j. , 2013 , arxiv e - prints , p. a. , seigar , m. s. , 1999 , 350 , 791 , t. , ellis , r. , jullo , e. , richard , j. , 2010 , apjl , 725 , l176 , j. & briggs , f. , 1992 , a&a , 1992 , 235 , 335 kennicutt , jr .
, r. c. , lee , j. c. , funes , jos g. , s. j. , sakai , s. , akiyama , s. , 2008 , apjs , 178 , 247 , jr . ,
r. c. , tamblyn , p. , congdon , c. e. , 1994 , apj , 435 , 22 , m. , prugniel , p. , bouchard , a. , wu , y. , 2009 , 501 , 1269 , m. , prugniel , p. , de rijcke , s. , 2008 , astronomische nachrichten , 329 , 968 , m. , prugniel , p. , ocvirk , p. , le borgne , d. , soubiran , c. , 2008 , mnras , 385 , 1998 kormendy j. , 1979 , apj , 227 , 714 kormendy j. , 1981 , in structure and evolution of normal galaxies , fall s.m .
, lynden - bell d. , eds . , pp.85 - 110 kormendy j. , 1982 , apj , 257 , 75 , j. , kennicutt , jr . , r. c. , 2004 , ara&a , 42 , 603 , h. , 2000 , mnras , 315 , 184 kuntschner h. , et al .
, 2006 , mnras , 369 , 497 , s. , shlosman , i. , knapen , j. h. , peletier , r. f. , 2002 , apj , 567 , 97 , 1985 , aj , 90 , 2652 , r. b. , 1976 , mnras , 176 , 31 , j .- f . ,
bruzual , g. , pell , r. , lanon , a. , rocca - volmerange , b. , sanahuja , b. , schaerer , d. , soubiran , c. , vlchez - gmez , r. , 2003 , a&a , 402 , 433 , j. r. d. , acharova , i. a. , mishurov , y. n. , 2003 , apj , 589 , 210 , m. , cabrera - lavers , a. , garzn , f. , hammersley , p. l. , 2002 , a&a , 394 , 883 macarthur , l. a. , courteau , s. , bell , e. , holtzman , j. a. , 2004 , apjs , 152 , 175 , l. a. , gonzlez , j. j. , courteau , 2009 , mnras , 395 , 28 maciel w. j. , costa r .d .
d. , 2013 , arxiv e - prints , w. j. , costa , r. d. d. , uchida , m. m. m. , 2003 , a&a , 398 , 667 , c. , strmbck , g. , 2011 , mnras , 418 , 2785 , p. and roy , j .- r . , 1995 , apj , 445 , 161 , i. , shlosman , i. , 2004 , apjl , 613 , l29 , h. j. , mao , s. , white , s. d. m. , 1998 , mnras , 295 , 319 , m. , daz , a. i. , 2005 , mnras , 358 , 521 , b. k. , holtzman , j. a. , 2006 , mnras , 371 , 583 , j. , 2005 , a&a , 430 , 95 , j. c. , gil de paz , a. , boissier , s. , zamorano , j. , jarrett , t. , gallego , j. , madore , b. f. , 2007 , apj , 658 , 1006 , t. , ostriker , j. p. , 2006 ,
mnras , 366 , 899 , g. , pedichini , f. , righini , m. , 1992 , a&a , 256 , 79 , p. , pichon , c. , lanon , a. , thibaut , e. , 2006 , mnras , 365 , 74 , p. , pichon , c. , lanon , a. , thibaut , e. , 2006 , mnras , 365 , 46 , b. , heavens , a. f. , jimenez , r. , 2003 , mnras , 343 , 1145 , i. , snchez - blzquez , p. , zurita , a. , 2009 , a&a , 495 , 775 pilkington k. , et al . , 2012 , a&a , 540 , a56 , m. and trujillo , i. , 2006 , a&a , 454 , 759 , j. et al . , 2012 , a&a , 539 , a93 , n.,sellwood , j. a. , james , r. a. , kahn , f. d. , 1991 , nature , 351 , 411 , m. s. , 1962 , aj , 1962 , 67 , 437 , j. c.,courteau , s.,snchez - blzquez , p.,mcdonald , m. , 2012 , apj , 758 , 41 , f. f. , 2011 , new astronomy , 16 , 220 , f. f. , kennicutt , r. c. , snchez , s. f. , daz , a. i. , pasquali , a. , johnson , b. d. , hao , c. n. , 2010 , mnras , 405 , 735 , r. , debattista , v. p. , quinn , t. r. , stinson , g. s. , wadsley , j. , 2008 , apjl , 684 , l79 , j .- r . and walsh , j. r. , 1997 , mnras , 288 , 715 , d. s. n. , kewley , l. j. , barnes , j. e. , 2010 , apjl , 710 , l156 , f. k. and smirnov , m. a. , 2004 , astronomy reports , 48 , 995 snchez s.f .
et al . , 2012 ,
a&a , 538 , a8 , s. f. , rosales - ortega , f. f. , kennicutt , r. c. , johnson , b. d. , diaz , a. i. , pasquali , a. , hao , c. n. , 2011 , mnras , 410 , 313 snchez - blzquez , p. , courty , s.,gibson , b. k. , brook , c. b. , 2009 , mnras , 398 , 591 , 2006c , a&a , 457 , 823 , p. , gorgas , j. , cardiel , n. , gonzlez , j. j. , 2006a , a&a 457 , 787 , 2006b , a&a , 457 , 809 , p. , ocvirk , p.,gibson , b. k.,prez , i. , peletier , r. f. , 2001 , mnras , 415 , 709 snchez - blzquez et al . , 2006d , mnras , 371 , 703 , m. , rix , h .- w . ,
shields , j. c. , ho , l. c. , barth , a. j. , rudnick , g. , filippenko , a. v. , sargent , w. l. w. , 2005 , apj , 628 , 169 , s. and lpine , j. r. d. , 2013 , mnras , 428 , 625 , jr . , s. and lpine , j. r. d. and marcon - uchida , m. m. , 2011 , mnras , 412 , 1741 , m. s. , 2002 , a&a , 393 , 499 , j. a. , binney , j. j. , 2002 , mnras , 336 , 785 , p. and trager , s. c. , 2007 , mnras , 374 , 769 , j. , sellwood , j. a. , 2004 , apj , 604 , 614 sheth et al . , 2008 , apj , 675 , 1141 , g. s. , van der kruit , p. c. , 1984 , a&a , 132 , 20 , o. k. , afanasiev , v. l. , 2002 , a&a , 385 , 1 , r. j. , lucey , j. r. , hudson , m. j. , 2007 , mnras , 381 , 1035 , l. , haywood , m. , 2010 , apj , 714 , 1096 , d. , maraston , c. , bender , r. , mendes de oliveira , c. , 2005 , apj , 621 , 673 , r. , heavens , a. f. , jimenez , r. , panter , b. , 2007 , mnras , 381 , 1252 , s. c. , faber , s. m. , worthey , g. , gonzlez , j. j. , 2000 , aj , 120 , 165 , s. c. , worthey , g. , faber , s. m. , burstein , d. , gonzalez , j. j. , 1998 , apjs , 116 , 1 , f. , gupta , r. , rose , j. a. , singh , h. p. , bell , d. j. , 2004 , apjs , 152 , 251 vazdekis a. , 1999 , apj , 513 , 224 , a. , cenarro , a. j. , gorgas , j. , cardiel , n. , peletier , r. f. , 2003 , mnras , 340 , 1317 , a. , snchez - blzquez , p. , falcn - barroso , j. , cenarro , a. j. , beasley , m. a. , cardiel , n. , gorgas , j.,peletier , r. f. , 2010 , mnras , 404 , 1639 , m. b. , edmunds , m. g. , 1992 , mnras , 259 , 121 , b. p. , adler , d. s. , 1995 , aj , 109 , 134 , s. d. m. , frenk , c. s. , 1992 , apj , 379 , 52 wozniak , h. , 2007 , a&a , 465 , 1 , p. , dalcanton , j. j. , 2008 , apj , 683 , 707 , p. , mccommas , l. p. , dalcanton , j. j. , williams , b. f. , 2009 , aj , 2009 , 137 , 469 , p. , rokar , r. , debattista , v. p. , 2012 ,
apj , 752 , 97 , t .-
kewley , l. j. , swinbank , a. m. , richard , j. , livermore , r. c. , 2011 , apjl , 732 , l14 , h. j. and bresolin , f. , 2011 , aj , 141 , 192 , d. , 1992 , apjl , 390 , l73 , h. , zhang , w. , yang , y. , zhou , x. , jiang , z. , ma , j. , wu , z. , wu , j. , zhang , t. , fan , z.
it is known that the measurement of the velocity dispersion depends on the metallicity of the stellar template ( laird & levison 1985 ; bender 1990 ; koleva , prugniel & de rijcke 2008 ) .
a deeper absorption feature can be obtained by either increasing the metallicity of decreasing the broadening , which leads to degeneracies when trying to deriving both parameters at the same time . to quantify the differences in the derived parameters with and without fixing the kinematics , we have run steckmap allowing to fit both ,
the radial velocity and velocity dispersion of the voronoi - binned spectra in ngc 628 and compare the derived , mean parameters , with those obtained when fixing the kinematics ( that is the approach followed in the main paper ) .
the results are shown in fig .
[ compara_kin ] . as can be seen , while the mean ages are not strongly affected by fixing or not the kinematics
, the metallicities can be by as much as @xmath0 1 dex .
the net effect is that , without fixing the kinematics , the metallicities tend to be lower .
furthermore , there is a trend for which the differences increase with the mean metallicity of the population .
we showed in snchez - blzquez et al .
( 2011 ) that this effect is not exclusive of our employed technique to derive star formation histories , but that it is also seen if other codes , as starlight is used .
the differences , also depend on the signal - to - noise . to further emphasize this
, we also include some spectra resulting from a voronoi - binned spectra imposing a lower signal - to - noise cat ( showed with different symbols in the figure ) .
it can be seen that the lower the signal - to - noise , the larger the differences .
figure [ fit_cont ] shows the comparison of our derived mean stellar age and metallicity ( weighted with both , light and mass ) and that obtained when we allow for the continuum shape to be fit .
we have use the spectra of the voronoi - binned data . as can be seen in the figures , except for the luminosity - weighted age , the rest of the parameters
are strongly affected by fitting or not the continuum .
it is also clear that for the those spectra with the highest signal - to - noise ( dark colored symbols ) , the differences are much lower than for the spectra with the lowest signal - to - noise ( light color and crosses . )
the maps for all the lick / ids indices are available in the electronic edition of the paper .
the indices have been measured in the voronoi - binned spectra broadened to the total resolution of 14(fwhm ) , that is , matching the lis-14 spectral system defined in vazdekis et al.(2010 ) .
we measured all the lick / ids indices with the definitions of trager ( 1998 ) and the 4000-break with the definition given in bruzual et al . | we present a stellar population analysis of the nearby , face - on , sa(s)c galaxy , ngc 628 , which is part of the ppak ifs nearby galaxies survey ( pings ) .
the data cover a field of view of @xmath0 6 arcmin in diameter with a sampling of @xmath02.7 arcsec per spectrum and a wavelength range ( 3700 - 7000 ) .
we apply spectral inversion methods to derive 2-dimensional maps of star formation histories and chemical enrichment .
we present maps of the mean ( luminosity- and mass - weighted ) age and metallicity that reveal the presence of structures such as a nuclear ring , previously seen in molecular gas .
the disk is dominated in mass by an old stellar component at all radii sampled by our data , while the percentage of young stars increase with radius .
the mean stellar age and metallicity profiles have a two defined regions , an inner one with flatter gradients ( even slightly positive ) and an external ones with a negative , steeper one , separated at @xmath060 arcsec .
this break in the profiles is more prominent in the old stellar component .
the young component shows a metallicity gradient that is very similar to that of the gas , and that is flatter in the whole disc .
the agreement between the metallicity gradient of the young stars and the gas , and the recovery of the measured colours from our derived star formation histories validate the techniques to recover the age - metallicity and the star formation histories in disc galaxies from integrated spectra .
we speculate about the possible origin of the break and conclude that the most likely scenario is that we are seeing , in the center of ngc 628 , a dissolving bar , as predicted in some numerical simulations .
[ firstpage ] galaxies : abundances ; galaxies : evolution ; galaxies : formation ; galaxies : spiral ; galaxies : stellar content |
the ngc 1023 group is one of the most extensively systems studied ( 300 related references are listed in the nasa / ipac extragalactic database ) .
the first large scale study of it has been performed by @xcite .
later on the presence of several outstanding dwarf galaxies drew the attention of @xcite who studied in detail four of them in the optical bands .
the most peculiar feature of ngc 1023 is its proximity to the ( likely tidally interacting ) fainter galaxy ngc 1023a .
the latter galaxy appears as a small companion located at the east end of ngc 1023 .
it appears as a low - luminosity condensation partially embedded in the disc of the larger galaxy .
this configuration gives the appearance of an asymmetric spiral arm and led @xcite to include the two galaxies in his atlas of peculiar galaxies as arp 135 .
only later @xcite proposed ngc 1023a to be an interacting individual galaxy stressing also its coincidence with a cloud of a neutral hydrogen ( hi ) identified by @xcite .
additional , more recent hi observations confirming the sancisi s results ( presence of tails and bridges ) has been performed by morganti et al .
the separation ( and different stellar population ) of the two objects is assured today also by the uv images of the galex satellite ( morrissey et al .
in particular ngc 1023a appears to be an irregular galaxy close to the sb0 ngc 1023 .
moreover the surface brightness , fuv and nuv galex profiles of ngc 1023 are included in the ultraviolet atlas of nearby galaxies of gil de paz et al . ( 2007 ) .
current programs ( e.g. trentham & tully 2009 ) are intended to explore the ngc 1023 environment in order to detect the faintest candidate dwarfs ( and establish their possible membership to the group ) ; the outcome of such research could indeed help to clarify , for instance , the so called fundamental missing satellite problem .
moreover ngc 1023 looks to be the first known s0 for which a proper kinematical approach based on pne ( cortesi et al .
2011 ) has shown that it likely originates from the secular evolution of a parent spiral galaxy instead of a more disruptive fast merger . in this paper
we analyze uv data in the central region of the group and we use them to characterize its core with sph simulations .
these are used also to investigate the dark matter distribution in the group environment as not confined in the haloes of the single galaxies .
the work is structured as follows : section 2 gives a detailed description of the dominant galaxy , ngc 1023 , while section 3 provides an exhaustive description of the uv photometric analysis of both ngc 1023 and ngc 1023a .
the adopted sph modeling is described in section 4 .
the conclusions are finally listed in section 5 .
ngc 1023 is an almost edge - on lenticular galaxy , classified as sb0 by most catalogues ( e.g. rc3 ; de vaucouleurs et al .
1991 ) , at an estimated distance of 11.4 mpc ( tonry et al . 2001 ) .
it belongs to a sparse group , rich in spirals , and is the brightest member of the lgg 70 group ( garcia 1993 ) . in the optical images ( the carnegie atlas 1994 ) the galaxy appears to be composed of a faint stellar disk , a inner and brighter lens and a bar at @xmath045@xmath1 from the apparent disk major axis .
a central nuclear disk is also present , with relatively young stars ( cf .
debattista et al .
2002 ) . from optical line index measurements ( boroson et al .
2011 , their table 1 ) derive a velocity dispersion of 210km / s and a stellar age of 4.7 gyr .
the redshift difference with the companion ngc 1023a is @xmath2v=127@xmath330 km s@xmath4 ( capaccioli et al .
1986 ) and the projected distance @xmath07.6 kpc ( barbon & capaccioli 1975 ; adopted h@xmath5 = 75 km s@xmath4 mpc@xmath4 ) . in bluer and deeper images the smaller galaxy exhibits some luminous knots that are consistent with a late - type dwarf galaxy .
we used fuv and nuv background - subtracted , de - archived intensity images from the galex pipeline . in order to estimate the uncertainties of uv ( ab ) magnitudes , we propagated poisson statistical errors on source and background counts .
in addition to the statistical error , we added an uncertainty to account for systematic inaccuracies in the zero - point of the absolute calibration of 0.05 and 0.03 mag for fuv and nuv respectively ( morrissey et al .
2007 ) .
fuv and nuv magnitudes have been computed as @xmath6 where cr is the dead - time corrected , flat fielded count rate , and the zero points are zp=18.82 and zp=20.08 in fuv and nuv respectively ( morrissey et al . 2007 ) .
we corrected uv magnitudes by adopting a foreground reddening from schlegel et al .
( 1998 ) for both ngc 1023 and ngc 1023a , namely e(b - v)=0.061 . with reference to the extinction curve of savage & mathis ( 1979 )
, this translates into the extinction values a@xmath7=0.56 and a@xmath8=0.49 respectively , adopting @xmath9=2267 and 1516 for the two above filters . for our study
, first we derive the photometic parameters of ngc 1023 and ngc 1023a and as a second step we cross - correlate the @xcite sample of objects with the galex photometric catalogue .
we found 9 matches ( together with ngc 1023 and ngc 1023a ) over 28 objects .
all the galaxies with know redshift are detected in uv bands while only two ( namely ngc1023 - 15 and ngc1023 - 17 ) of the dwarfs listed in @xcite have uv emission .
in addition to these galaxies we discovered two new objects , identified as c1 and c2 .
our galex identification of the galaxies , as listed in @xcite and @xcite , are shown in fig .
[ galex_field ] . both fuv and nuv surface photometry
have been performed using iraf stsdas ellipse routine .
ellipse computes a fourier expansion for each successive isophote ( jedrzejewski 1987 ) , resulting in the surface photometric profiles .
all corrected magnitudes m@xmath10s of ngc 1023 , ngc 1023a , and davies & kinman s and @xcite objects are given in table [ tab : fos_indb ] . in figures
[ galex_field1],[galex_field2 ] we show the nuv , fuv and poss2 r images of all the objects we have detected ( in nuv or fuv ) .
we computed the ultraviolet integrated photometry ( fuv@xmath11nuv ) of each uv - bright source and the color profiles of the two brightest galaxies , namely ngc 1023 and ngc 1023a ( fig [ col1 ] ) .
integrated photometric properties are given in table [ tab : fos_indb ] while photometric parameters for the fits are listed in table [ re ] .
errors in total magnitudes were obtained from the correlation coefficient of the model fit vs. the observed values .
we fitted the luminosity profiles of ngc 1023 assuming two - components : a bulge following an r@xmath12 law , and an exponential disc .
the inner regions of the profiles , affected by the galex psf ( @xmath0 5@xmath13 ) , were excluded in performing interpolation .
the lens and bar are not visible in uv images and the ellipse interpolations find constant ellipticity and p.a . values .
in the case of ngc 1023 they are @xmath140.23 ( nuv ) and 0.12 ( fuv ) ; p.a@xmath080@xmath1 ( nuv ) and 83@xmath1 ( fuv ) .
the values of the bulge effective radii in fuv and nuv reported in table [ tab : fos_indb ] are quite similar and are in agreement with those found by optical photometry in r band , reported in the work of @xcite : r@xmath15=32 `` .
the stellar disk , when observed in red light , appears larger , with an r@xmath16=62 '' compared to values lower than 50 " found in the uv .
the fuv - nuv color profile ( top panel of fig .
[ col1 ] ) indicate that the nucleus is bluer that the external regions .
this is confirmed also by the nuv - r color profile plotted in fig .
[ col ] together with the fuv - r .
the r data were derived from @xcite .
both uv profile components can be fitted in ngc 1023a with an exponential disk .
the routine ellipse gives almost constant ellipticity and position angles .
they are : @xmath170.45 ( nuv ) and 0.38 ( fuv ) , and p.a@xmath016@xmath1 ( both in nuv and fuv ) .
the effective radii and surface brightness are listed in table [ re ] together with the total magnitudes of both galaxies .
integrated photometric properties are given in table [ tab : fos_indb ] while photometric parameters for the fits are listed in table [ re ] .
contrary to what observed in ngc 1023 the color profile in ngc 1023a is constant around a value @xmath00.8 .
we measured the uv magnitudes of the two uv emission regions , labeled c1 and c2 ( see fig .
[ galex_field ] , [ galex_field1 ] , [ galex_field2 ] ) inside an aperture of r=15 arcsec ; the values are listed in table [ tab : fos_indb ] . in fig .
[ colori ] we plot the colors of the main objects in the group .
we note that ngc 1023 and ngc 1023a have similar colors , whereas the `` dwarfs '' ngc 1023b , ngc 1023d , ugc 2157 and especially ngc 1023c are quite bluer in the uv ( both in fuv - nuv and nuv - b ) .
ugc 2165 on the contrary is very red , in particular the fuv flux measured can be considered as an upper limit and may indicate a very old stellar population . [
cols="<,^,^,^,^,^,^,^,^,^,^ " , ] the present - day star formation rate ( sfr ) of each uv source can be derived following @xcite using its uv continuum luminosity and the relation + sfr@xmath8 ( m@xmath18 yr@xmath4 ) = 1.4 @xmath19 10@xmath20l@xmath8(ergs s@xmath4 hz@xmath4 ) .
+ such a relation holds for a sapleter s imf with lower and upper mass limits 0.1m@xmath18 and 100m@xmath18 respectively .
we used our total fuv magnitudes to derive the sfrs which are given in table [ tab : fos_indb ] . using the previous equation we found that the sfr is not null in ngc 1023 although it provides very low values ( 2.68@xmath21m@xmath18 yr@xmath4 ) .
the spitzer data for the inner regions of this galaxy @xcite indicate no signature of sf .
the sfr in the other dwarf galaxies and in the two blobs ( c1 and c2 ) is even lower , with the lowest value , 0.01@xmath21m@xmath18 yr@xmath4 , for ugc 2165 .
however their colors ( see fig . [ colori ] ) are bluer than those of the two brighter galaxies of the group .
with the aim to match the overall seds and the global properties of ngc1023 and ngc1203a in a consistent way with the dynamical properties of the group , we select , from a large set of sph simulations , one case which gives us insights into the evolution of such a system and the group itself .
our sph simulations of galaxy formation and evolution are starting from the same initial conditions as described in @xcite ( mc03 in the following ) and @xcite i.e. , collapsing triaxial systems initially composed of dark matter ( dm ) and gas in different proportions and different total masses .
all the simulations include self gravity of gas , stars and dm , radiative cooling , hydrodynamical pressure , shock heating , artificial viscosity , star formation ( sf ) and feedback from evolved stars and type ii sne , and chemical enrichment as in mc03 ( and references therein ) .
they provide the synthetic sed at each evolutionary step .
the sed accounted for chemical evolution , stellar emission , internal extinction and re - emission by dust in a self - consistent way , as described in @xcite and @xcite ; this extends over four order of magnitude in wavelength , i.e. , from 0.1 to 1000 @xmath22 m .
so , each simulation self - consistently provides dynamic , morphological , and chemo - photometric evolution .
the initial mass function ( imf ) is of salpeter type with upper mass limit 100@xmath23 and lower mass limit 0.01@xmath23 ( salpeter , 1955 ) ( see @xcite and mc03 for a discussion ) .
all the model parameters here adopted had been tuned in previous papers where the integrated properties of simulated galaxies , i.e. , colors , absolute magnitudes , mass to luminosity ratios provided by different choices of model parameters after 15 gyr , had been successfully compared with those of local galaxies ( see also @xcite ; @xcite ) . in particular , the imf choice here provides a slightly higher sfr compared with the other possibilities examined , allows for the lowest feedback strength , and the expected rotational support when disk galaxies are formed .
moreover , as pointed out by @xcite , its slope is almost the same as the universal mass function which links imf of galaxies and stars to those of brown dwarfs , planets and small bodies ( meteoroids , asteroids ) @xcite . with respect to mc03 , the initial particle resolution here
is enhanced to more than 6@xmath1910@xmath24 ( see below ) instead of 2@xmath1910@xmath24 , and the time separation between the snapshots halved , 151myr .
a new large set of galaxy encounters involving systems with 1:1 mass ratios and different mass ratios have been performed with the same initial conditions as described in mc03 ( and references therein ) . by seeking to exploit a vast range of orbital parameters
, we carried out different simulations for each couple of interacting systems varying the orbital initial conditions in order to have , for the ideal keplerian orbit of two mass points of mass equal to 10@xmath25 , or 10@xmath26 , the first peri - center separation , p , equal to the initial length of the major axis of the dm triaxial halo , i.e. 88 kpc for 10@xmath27 , or equal to 1/10 , 1/5 , and 1/3 of the same axis for 10@xmath26 encounters .
for each of these separations , we changed the eccentricity in order to have hyperbolic orbits of different energy .
the spins of the systems are equal ( mc03 ) , generally parallel each other , and perpendicular to the orbital plane , so we studied direct encounters .
however , some cases with misaligned spins have been also analyzed in order to deepen the effects of the system initial rotation on the results .
moreover , for a given set of encounters with the same orbital parameters we also examined the role of increasing initial gas fractions .
all these new simulations will be fully discussed elsewhere ( mazzei , in prep ) .
a simulation able to match the global properties of ngc 1023 is chosen and discussed below .
all the observables , i.e. absolute magnitudes , colors , which correspond to the sed over almost four order of magnitude in wavelength , and mass - to - light luminosity ratios are self - consistently derived from our simulation at the snapshot selected , and successfully compared with the available data .
the whole sed of ngc 1023 and its global properties are well matched by a simulation which provides a minor merger .
the total mass of the primary system is 1@xmath1910@xmath28 m@xmath18 with gas fraction 0.01 , that - one of the companion is 5 times less with the same gas fraction .
so , the starting point is given by two triaxial collapsing systems with total mass 1.2@xmath1910@xmath28 m@xmath18 , and 60 times less of gas , i.e. 2@xmath1910@xmath29 m@xmath18 .
the mass particle resolution is 5@xmath1910@xmath30 m@xmath18 for gas and 4.95@xmath1910@xmath31 m@xmath18 for dm particles .
this requires 63838 initial particles .
the triaxiality ratio , @xmath32 @xcite , is 0.84 .
+ the first pericenter separation , 433 kpc , corresponds to 1/3 of the major axis of the primary system ; the orbit eccentricity is 1.3 .
and the anomaly corresponds to 200 degrees .
the spins of the systems are equal ( @xmath33=0.06 , mc03 ) , parallel , so we are dealing with a direct encounter , and both aligned with the shorter of their principal axes .
+ stars are born in the inner regions of their halos after 3gyr from the beginning .
galaxies grow changing their shapes step by step , as their trajectories are approaching and their halos mixing .
the fit corresponds to a galaxy age of 11.9 gyr .
the strong interaction phase between the systems started 7.7gyr ago and the final merger 1.8gyr ago .
figure [ accgas ] shows the gas accretion history inside a sphere centered on the luminous ( v band ) centre of the galaxy system with a radius of 50kpc . the mass of gas with temperature @xmath3410000k , which represents the upper limit of the cold gas mass inside the selected radius ( its cooling time scale is very shorter than the snapshot time range , 151myr ) , is 9.50@xmath1910@xmath35 m@xmath36 at the age of our fit ( the vertical line in fig .
[ accgas ] ) .
such a value well agrees with that derived from eq .
( 1 ) of @xcite using d=5.93mpc ( ned from 3k cmb ) and fi , the flux integrated over line , 80jykm / s @xcite , i.e. , 7.3@xmath1910@xmath35 m@xmath36 ; this value is at least a factor 3.7 more using the distance in ( * ? ? ?
* their table 2 , 11.4 mpc ) .
figure [ mdist ] shows the mass distribution at the selected snapshot inside 50kpc , corresponding to 15r@xmath37 of ngc 1023 .
the b band absolute magnitude of the model , @xmath38 , agrees with the value in table 1 of @xcite and with table 1 of ( * ? ? ?
* and references therein , @xmath39 ) ; r band absolute magnitude , @xmath40 , agrees with the estimate of @xcite , @xmath41 ( their table 2 ) . furthermore , all the available data , extending over three order of magnitude in wavelength , are well matched by the sed provided by our simulation , as shown in fig .
[ modsed ] .
the fir region is not strongly constrained by the data , since upper limits of 25 , 60 and 100@xmath22 m fluxes are only available ; dust components ( warm and cold with pah as discussed in @xcite ) with the same average properties as derived by @xcite and @xcite for a complete sample of nearby early - type galaxies are included ( r@xmath42=100r@xmath43 , @xcite ) .
figure [ morfo_ii ] shows r band maps of different projections of the selected snapshot .
figure [ morfo ] compares , on the same scale , the uv galex image with superimposed the hi map of ngc 1023 , with r map of xy projection as in fig .
[ morfo_ii ] .
star populations inside the previous maps are 5.0 gyr old in average , in agreement with @xcite ( see section 2 ) .
the luminosity weighted age of the model does not depend strongly on the radius nor on the band , it is 2.5 - 3 gyr going from the inner regions up to 50kpc . from the study of @xcite ,
the virial mass of the ngc 1023 group inside 315kpc ( their virial / harmonic radius ) , is 6.4@xmath33@xmath1910@xmath25 m@xmath44 , and l@xmath45=2.0@xmath1910@xmath46 l@xmath44 .
we predict a total mass of 5.54@xmath1910@xmath25 m@xmath44 with a stellar mass of 1.20@xmath1910@xmath29 m@xmath44 , and l@xmath47=2.56@xmath1910@xmath46 l@xmath44 inside the same radius .
the instantaneous sfr in the inner region of the simulated galaxy agrees with sauron results by shapiro et al .
( 2010 ) who measure sfr@xmath00 @xmath48 yr@xmath4 ( see section 3.1 ) , moreover the total sfr at the selected snapshot , 0.045 m@xmath44 yr@xmath4 , is in well agreement with our estimates from fuv luminosity .
note that to compare these values with those in table [ tab : fos_indb ] a multiplying factor of @xmath02.4 has to be applied to the data in such a table to account for the lower mass limit of our simulation .
figure [ vel ] compares our predicted average velocity distribution of stars at each galaxy radius with the same from the data in fig . 4 of cortesi et al .
velocities from our simulations are higher than 7% ; such a reduction factor allow us to account for the inclination angle of ngc 1023 that is @xmath49 .
note that the green line in fig .
[ vel ] corresponds to the average velocity distribution of planetary nebulae ; we select their velocities along the major axis of such a galaxy ( pa=@xmath50 ) accounting for a recession velocity of 637km / s .
first , we reanalyzed the inner group ngc 1023 by means of deep galex images .
so we extend to uv range the detailed optical observations of both the four dwarfs studied by davies & kinman ( 1984 ) and the sample of @xcite .
these observations led to the discovery of two other very faint uv - emitting objects , possibly members of the group itself ( see fig .
[ galex_field2 ] and table [ tab : fos_indb ] ) .
all the five dwarfs show nuv - b colors bluer than those of the two brighter members of the group . in particular ugc 2165
is very faint in fuv showing extremely red fuv - nuv colors .
moreover , with the aim at matching the specific overall seds and the global properties of ngc1023 and ngc1203a in a consistent way with the dynamical properties of the group , we select , from a large set of sph simulations , one case which gives us insights into the evolution of such a system and the group itself .
such sph simulations of galaxy formation and evolution are starting from a collapsing triaxial systems initially composed of dark matter ( dm ) and gas in different proportions and different total masses .
these simulations allowed to investigate the dm distribution in the group environment finding that this is not confined to the halo of single galaxies .
the whole sed of ngc 1023 and its global properties are well matched by the simulation which provides a minor merger , 1:5 @xmath51 ) .
the merger phase started 1.8 gyr before the selected snapshot .
the whole outcome of the model is in strict agreement with @xcite results : their estimate of the the r band absolute magnitude is @xmath41 to be compared with our @xmath40 .
in addition from their study of the ngc 1023 group , the virial mass inside 315kpc ( their virial / harmonic radius ) , is 6.4@xmath33@xmath1910@xmath25 @xmath48 , and l@xmath47=2.0@xmath1910@xmath46 @xmath52 , according to our prediction of a total mass of 5.54@xmath1910@xmath25 @xmath48 with a stellar mass of 1.20@xmath1910@xmath29 @xmath48 , and l@xmath47=2.56@xmath1910@xmath46 @xmath52 inside the same radius .
looking at its evolution , we predict the group more loose and poor , still no - virialized when the final merger phase is not yet begun , 2 gyr before our fit , which corresponds to an age of 11.9 gyr .
b and r luminosities are almost the same as at the age of the fit , but there is 20% less mass within the same reference radius , i.e. 315 kpc @xcite .
furthermore , b and r luminosities reduce by 1 and 0.5 mag respectively 2 gyr after the age of our fit , the total mass being the same .
we thank the referee dr .
n. trentham whose comments greatly improved the paper .
we acknowledge the financial contribution from university of padova ( ex 60% ) and the agreement asi - inaf i/009/10/0 .
galex is a nasa small explorer , launched in april 2003 .
galex is operated for nasa by california institute of technology under nasa contract nas-98034 .
99 arp , h. 1966 , apjs , 14 , 1 barbon , r. , & capaccioli , m. 1975 , a&a , 42 , 103 bettoni d. , galletta , g. , rampazzo , r. , marino , a. , mazzei , p. & buson , l. m. 2011 , a&a , 534 , a24 binggeli , b. & hascher , t. 2007 , pasp , 119 , 592 boroson , b. , dong - woo , k. , & fabbiano , g. 2011 , apj 729 , 12 capaccioli , m. , lorenz , h. , & afanasjev , v. l. 1986 , a&a , 169 , 54 cappellari , m. , emsellem , e. , bacon , r. et al .
2007 , mnras , 379 , 418 cortesi , a. , merrifield , m. r. , arnaboldi , m. , et al .
2011 , mnras , 414 , 642 curir , a. , mazzei , p. 1999
, , new astron . , 4 , 1 davies , r. d. , & kinman , t. d. 1984 , mnras , 207 , 173 debattista , v. p. , corsini , e. m. , & aguerri , j. a. l. 2002 , mnras , 332 , 65 de vaucouleurs , g. , de vaucouleurs , a. , corwin , h. g. , jr . , et al . 1991 ( rc3 )
emsellem , e. , cappellari , m. , krajnovic , d. , et al .
2007 , mnras , 379 , 401 .
garcia , a. m. 1993 , a&as , 100 , 47 gil de paz , a. , boissier , s. , madore , b. f. , et al .
2007 , apjs , 173 , 185 jedrzejewski , r. i. 1987 , mnras , 226 , 747 jeong , h. , sukyoung , k. , bureau , m. et al .
2009 , mnras , 398 , 2028 kennicutt , r. c. , jr .
1998 , ara&a , 36 , 189 kennicutt , r. c. , jr .
1998 , ara&a , 36 , 189 knapp , g. , turner , e. , cunniffe , p.e 1985 , aj , 90 , 454 mazzei , p. 2003
sait , 74 , 498 mazzei , p. , xu , c. , & de zotti , g. 1992 , a&a , 256 , 45 mazzei , p. & de zotti , g. , xu , c. 1994 , apj , 422 , 81 mazzei , p. & de zotti , g. 1994 , apj , 426 , 97 mazzei , p. & curir , a. 2003 , apj , 591 , 784 : mc03 mazzei , p. 2004
, recent reaserch dev . in astron . & astrop . , 1 , 457 ed .
pandalai , s.g . , research signpost ; ( arxiv : astro - ph/0401509 )
morganti , r. , de zeeuw , p. t. , oosterloo , t. a. , et al .
2006 , mnras , 371 , 157 morrissey , p. , et al . 2007 ,
apjs 173 , 682 noordermeer , e. , merrifield , m. r. , coccato , l. , et al .
2005 , mnras , 384 , 943 pahre , m.p , ashby , m. l. n , fazio , g. g. , and willner , s. p. 2004 , apjss , 154 , 235 salpeter , e.e .
, 1955 , apj , 121 , 161 sancisi , r. , van woerden , h. , davies , r. d. , & hart , l. 1984 , mnras , 210 , 497 savage , b. d. , & mathis , j. s. 1979 , ara&a , 17 , 73 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , apj , 500 , 525 shapiro , k. l. , falcn - barroso , j. , van de ven , g. et al .
2010 , mnras , 402 , 2140 spavone , m. , iodice , e. , calvi , r. , bettoni , d. , galletta , g. , longo , g. , mazzei , p. & minervini , g. 2009 , mnras , 393 , 317 tonry , j. l. , dressler , a. , blakeslee , j. p. , et al .
2001 , apj , 546 , 681 trentham , n. , & tully , r.b . , 2009 , mnras , 398 , 722 tully , r. b. 1980 , apj , 237 , 390 warren , m. , quinn , p. , salmon , j. & zurek , w. 1992 , apj , 399 , 405 | the ngc 1023 group is one of the most studied nearby groups .
we want to give an insight into the evolution of its innermost region by means of ultraviolet observations and proper models .
we used the fuv and nuv galex archival data as well as a large set of sph simulations with chemo - photometric implementation . from the uv observations we found that several , already known , dwarf galaxies very close to ngc 1023
are also detected in uv and two more objects ( with no optical counterpart ) can be added to the group . using these data we construct exhaustive models to account for their formation .
we find that the whole sed of ngc 1023 and its global properties are well matched by a simulation which provides a minor merger with a companion system 5 times less massive .
the strong interaction phase started 7.7gyr ago and the final merger 1.8gyr ago .
[ firstpage ] galaxies : structure galaxies : individual : ngc 1023 |
the seminal work of anderson@xcite raising the possibility that disorder can lead to non - diffusive behavior ( the so called localized regime ) refocused the attention of the physics community on the problem of the propagation of waves in disordered systems . in the last two decades new theoretical ideas ( like the scaling theory of localization@xcite , weak localization@xcite , universal conductance fluctuations@xcite and wigner dwelling times@xcite ) were advanced , and a new field ( soon called mesoscopic physics ) emerged .
it reached and influenced many experimental areas , among them electronic systems@xcite , microwaves@xcite , optics@xcite , acoustics@xcite , geophysics@xcite , laser physics@xcite , medical physics@xcite and atomic physics@xcite .
it has become an extremely important problem in this field to understand what should be the signature of the propagation of a signal in the different regimes ( ballistic , diffusive , localized ) since concomitant phenomena , like absorption can complicate the interpretation of experimental results@xcite .
it is for that reason that theoretical analyses of the characteristics of the propagation , and in particular its statistical properties@xcite@xcite , are of great interests , since those properties have become recently experimentally accessible@xcite@xcite .
when the inelastic scattering length in a system is large compared to its size , the wave propagates coherently in the sense of its phase being preserved while its direction is randomized by elastic scattering processes with the impurities constituting the random medium .
it is natural , then , to establish connections between the coherent propagation of a wave and the statistical theory of random walks@xcite , where the continuum limit is known to lead to diffusive theory@xcite .
this connection is relevant to the propagation of an electronic current through certain layered nanostructures , since they can be modeled as a collection of random one - dimensional interfaces , through which a coherent signal can be transmitted or reflected while being scattered at each interface .
the connection between the effect of the disorder of the propagating medium and the statistical randomness of the mathematical description of the diffusion has not been satisfactorily clarified yet .
one aspect that did not receive enough attention in this approach is the fact that , since the scattering with the impurities is not isotropic and the cross section is normally enhanced in the forward direction , the statistical jumps are not independent from each other . in other words
the probability distribution of each step of these persistent random walks ( prw ) is dependent on the previous step@xcite .
the prw are more difficult to study than the standard random walks , where the probability distribution of each step is independent of what happened in the previous steps .
this can be seen from the fact that , for example , while a standard random walk can be easily mapped into a chain of non interacting spins ( in the presence of a magnetic field , if the random walk is biased ) , the same mapping for the prw leads to a chain of nearest - neighbor interacting spins ( the full ising model for the case of a 1-d prw)@xcite .
in particular in this work we will explore the connection between the propagation of a wave and the 1-d version of the prw called the t - r model@xcite , in which a particle moves with probability @xmath0 in the same direction as in the previous step and with probability @xmath1 reverses direction @xmath2 .
this type of model is more suited than the standard random walk model to explore the quasiballistic regime , which becomes more important in transport phenomena in nanostructures .
we will consider a model of the propagation of a wave first introduced for numerical purposes by edrei , kaveh and shapiro@xcite , and later applied extensively by vanneste _ _ et al__@xcite _ _ _ _ and sebbah@xcite .
each site is represented by a scattering matrix @xmath3 of dimension @xmath4 ( @xmath5 being the dimensionality ) connecting the @xmath6 incoming amplitudes at time @xmath0 , with the @xmath6 outgoing ones at time @xmath7 .
these amplitudes can be regarded as residing in the bonds ( 2 joining each site to its nearest neighbor , one in each direction ) .
these bonds represent free propagation between the sites , and their phase can be included in the @xmath8 matrix .
the wave function at time @xmath0 is , thus , determined by the complex numbers associated with each of the bonds at that time .
numerically this has the advantage of letting the stationary wave being built step by step in time , from an initial input without solving a huge diagonalization problem .
the matrices @xmath3 representing each scatterer are unitary and symmetric reflecting , respectively , the energy conservation and time reversal symmetry inherent to the problem .
if we adopt the simplifying assumption that the scatterers themselves are symmetric , these matrices simplify even further , since the following relations can be easily obtained for their elements @xmath9+(d-1)d^{2 } & = 0\nonumber\\ r\,d\,\cos[\varphi_{r}-\varphi_{d}]+t\,d\cos[\varphi_{t}-\varphi _ { d}]+(d-2)d^{2 } & = 0\nonumber\end{aligned}\ ] ] where @xmath10 and @xmath11 are the amplitudes and phases of the elements for reflection , transmission , and turning ( in 1d , @xmath12 ) .
the matrices @xmath3 have , thus , only two ( in 1d ) or three ( in d@xmath131 ) independent parameters . for example
, in 1d the scattering matrix representing one scatterer can be parameterized as : @xmath14{cc}% { \large r } & { \large i\,t}\\ { \large i\,t } & { \large r}% \end{array } \right ] \exp{\large [ i\,\varphi]\;\hspace{1in}(r}^{2}{\large + \;t}% ^{2})=1\label{1dmatrix}%\ ] ] in this simple version of the model the distances are measured in units of the mean free path ( the distance between scattering elements ) and the phase velocity of the wave is one .
the effect of the disorder in this model has been considered both analytically@xcite and numerically@xcite , by assuming a distribution of values for the variable @xmath1 , linked to the reflectivity of the scatterers .
alternatively , one can consider the effect of disorder in the variable @xmath15 , more closely linked to the distance between scatterers@xcite .
we will be interested in considering a signal incident from the left on a system composed of @xmath16 such scatterers , and in monitoring what is reflected or transmitted through it at different times , i.e. the output to the left of scatterer @xmath17 and to the right of scatterer @xmath16 , respectively .
it is important to observe that while calculating the response of the system to an applied pulse implies adding all the paths of a certain length ( that will interfere by arriving simultaneously to the boundary ) , the problem of determining the response to a continuous wave is simpler , since in the stationary state it is sufficient to add * all * the transmitting or reflecting paths as long as the phase factors associated with the propagation through each scatterer , @xmath18 $ ] , are included .
@xmath19 represents here the wavevector of the incident wave .
the amplitude factor for any particular path reaching the boundaries of the system can be easily written , and we can sum all the possible paths by performing a loop expansion .
for example , for a system of two scatterers the amplitudes associated with the simple paths illustrated in figure 1(a ) are , respectively , @xmath20)^{2}$ ] , @xmath20)^{2}(r\exp[i\,k])^{2}$ ] and @xmath20)^{2}(r\exp[i\,k])^{4}$ ] ; the total amplitude for transmission through the two scatterers , when all possible paths are considered , becomes then : @xmath21)^{2}}{1-(r\exp[i\,k])^{2}}\equiv\frac { ( i\,t\exp[i\,k])^{2}}{1-l_{1}}\label{t2wave}%\ ] ] for the three scatterers in figure 1(b ) , the amplitude factors associated with the paths shown are @xmath20)^{3}$ ] , @xmath20)^{3}% ( r\exp[i\,k])^{2}(i\,t\exp[i\,k])^{2}$ ] and @xmath22)^{3}% ( r\exp[i\,k])^{2}(i\,t\exp[i\,k])^{2}$ ] @xmath23)^{4}$ ] , respectively .
the bare loop of length 2 has an amplitude factor of @xmath24)^{2}(i\,t\exp[i\,k])^{2}$ ] , and when fully dressed by all the possible loops of length 1 becomes : [ ptb ] gq6c7p00.wmf @xmath25)^{2}(i\,t\exp[i\,k])^{2}}{\left ( 1-(r\exp[i\,k])^{2}\right ) ^{2}}\equiv\frac{l_{2}}{(1-\widetilde{l}_{1})^{2}}%\ ] ] we see that the factor associated with a bare loop of length @xmath26 is simply @xmath27)^{2}(i\,t\exp[i\,k])^{2(n-1)}$ ] . in terms of it
we can obtain a recursive expression for the general dressed loop of order @xmath26 @xmath28 @xmath29 it is possible now to add all the paths through a chain of @xmath16 scatterers , the amplitude of the transmission through the system , and express it in terms of the loops : @xmath30\;\;;\;\;t_{2}=\frac{(i\,t\,\exp[i\,k])^{2}% } { ( 1-\widetilde{l_{1}})}\;\;;\;\;t_{3}=\frac{(i\,t\,\exp[i\,k])^{3}% } { ( 1-\widetilde{l_{1}})^{2}(1-\widetilde{l_{2}})}%\ ] ] @xmath31)^{n}}{(1-\widetilde{l_{1}})^{(n-1)}% ( 1-\widetilde{l_{2}})^{(n-2)} ..... (1-\widetilde{l}_{n-1})}\label{wavetn}%\ ] ] as expected , @xmath32 is proportional to @xmath33 . a very similar expansion yields , for the reflection amplitudes : @xmath34\;\;;\;\;r_{2}=r_{1}+\frac{(i\,t\exp[i\,k])^{2}% \,(r\exp[i\,k])}{(1-\widetilde{l_{1}}\,)}%\]]@xmath35)^{4}\,(r\exp[i\,k])}{(1-\widetilde { l_{1}}\,)^{3}(1-\widetilde{l_{2}}\,)}%\ ] ] @xmath36)^{2(n-1)}(r\exp[i\,k])}{(1-\widetilde { l_{1}})^{(2n-3)}(1-\widetilde{l_{2}})^{(2n-5)} ..... (1-\widetilde{l}% _ { n-2})^{3}\,(1-\widetilde{l}_{n-1})}\label{wavern}%\ ] ] the expression for the amplitude of reflection through the system admits a clear interpretation in terms of a sum of all the paths that penetrate up to a certain depth in the system : @xmath37corresponds to the paths that do not enter into the system at all , @xmath38 represents the paths that penetrate up to a depth @xmath17 , @xmath39 are the paths that penetrate up to a depth @xmath40 , etc .
the connection of this expansion for the transmission and reflection amplitudes of the propagating wave with the prw is provided by the fact that we can write exactly the same expansion in loops in the t - r model , for all the possible paths of a random walker leading to transmission or reflection . here
transmission means the random walker arriving to a point further to the right of position @xmath16 , and reflection the walker reaching a point to the left of position @xmath17 .
in particular , if we ignore the phase factors @xmath18 $ ] and the imaginary unit in front of @xmath0 , the amplitude factors discussed above become the probabilities of the corresponding prw paths , and equations ( [ ldressedn]),([wavetn ] ) and ( [ wavern ] ) describe the dressed loops , probability of transmission and probability of reflection in the case of a prw as well , since for a random walk the * total * probability of transmission ( reflection ) is simply the sum of the probabilities of each of the paths leading to transmission ( reflection ) .
of course in each case , for a prw the connection between @xmath0 and @xmath1 is not provided by ( [ 1dmatrix ] ) , but by the simple expression @xmath41 .
it is easy to show now that the probability of transmission for a prw becomes , then : @xmath42 and the probability of reflection becomes simply @xmath43 .
identifying @xmath16 with the length of the system @xmath44 , and taking into account that in this model @xmath45 , we can see that in the limit of large @xmath16 the transmission tends to the diffusive limit @xmath46 . in the case of a wave ,
the transmission or reflection probabilities are given by the modulus square of the expressions calculated in ( [ wavetn ] ) and ( [ wavern ] ) .
this implies sums of the type : @xmath47 it is easy to see that , when adding the transmitting paths ' amplitudes , if we consider the contribution of only the first term in the expression above ( neglecting the correlations between the paths ) we will reobtain ( [ tnmrw ] ) , where the roles of @xmath0 and @xmath1 are played by the transmission coefficient @xmath48 , and the reflection coefficient @xmath49 .
it is precisely the overlap between paths ( the second term ) that leads to corrections to the diffusion theory results .
in the case of the prw , knowing the probabilities @xmath50and the lengths @xmath51of all the paths both in transmission and reflection , we are now in a position to study some of their statistical properties . defining a generating function @xmath52 , where we sum over all the transmitted paths , and a similar definition for @xmath53 where the sum is over all reflected paths , we can calculate the average length of the paths as @xmath54 obtaining for the average length : @xmath55[1-r]}\label{lavertr}%\ ] ] @xmath56 in the diffusive limit discussed above , these expressions reproduce the well known results of diffusive theory@xcite : @xmath57 .
analogously , @xmath58 , independent of the mean free path .
the expressions ( [ lavertr ] ) and ( [ laverref ] ) are exact , within the model , and remain valid in the diffusive as well as the ballistic limit @xmath59 , and can be used to interpolate between them .
this can be of interest in any situation in which the ballistic paths play an important role as in medical imaging . in a similar way one can express in terms of f(s ) the variance of the path length distribution , in transmission or reflection @xmath60^{2}%\ ] ] obtaining the expressions : @xmath61}{(1-r)^{2}[1+(n-1)\,r]^{2}}\label{vartran}%\ ] ] @xmath62^{2}% } \label{vrf}%\ ] ]
@xmath63 is @xmath64 here .
these exact expressions , which again allow us to extrapolate between the diffusive and ballistic regimes , in the diffusive limit tend to @xmath65 and @xmath66 respectively . in the case of transmission
the variance scales with the size of the system as the square of the average length ( non - self averaging property ) while in the case of reflection it scales even faster than it , exhibiting the well known noisy character of reflection@xcite .
figures 2 , 3 and 4 exhibit the results of numerical simulations of a persistent random walker on a system of size @xmath67 .
[ ptb ] gq6c7p01.wmf [ ptbptb ] gq6c7p02.wmf [ ptbptbptb ] gq6c7p03.wmf the transmission and reflection probabilities , the average length of the transmitted and reflected paths and the variance for the path lengths in transmission and reflection are represented as a function of the transmission parameter @xmath68 in all cases the agreement with the expression [ tnmrw ] and ( [ lavertr ] ) to ( [ vrf ] ) is excellent .
an interesting problem is how to define the probabilistic weights to perform a similar statistical analysis of the length of the paths in the case of a _ wave _ , since the amplitudes are complex numbers with a phase , and the different paths interfere with each other .
it makes sense to use the probabilities ( in the sense of the moduli of the complex amplitudes squared ) as those probabilistic weights .
as the discussion at the end of section [ loop ] shows , those weights describe the diffusive behavior of the system . adopting this ansatz
we can define again the generating function as in the case of the prw , and we can reobtain all the results in expressions ( [ lavertr ] ) to ( [ vrf ] ) , again replacing @xmath0 and @xmath1 by the transmission coefficient @xmath69 and the reflection coefficient @xmath70 , respectively .
all this analysis and the expansion in terms of loops can be easily generalized to a system in a higher number of dimensions . to fix ideas
let us consider , instead of a row of @xmath16 scatterers , a strip of @xmath71 rows of @xmath16 scatterers each , where the wave ( or the random walker ) is incident from left . for this system ,
the amplitudes of transmission or reflection through _ one _ column become ( instead of @xmath0 and @xmath1 ) the @xmath72 matrices @xmath73 and @xmath74 , whose element @xmath75 is the amplitude for a walker incident from the left on row @xmath76 to exit to the right on row @xmath77 , and an analogous definition for @xmath78 .
these amplitudes are obtained by adding all the possible paths that lead from the input @xmath76 to the output @xmath77 and are confined to _
one _ single column of the medium .
it is , then , not difficult to prove that the expansion in loops ( [ ldressedn ] ) remains valid as long as we replace in it @xmath0 and @xmath1 by these matrices , and all the products are understood as products of matrices .
similarly , the transmission @xmath79 and the reflection @xmath80 through the strip of length @xmath16 are obtained through the same replacement from the expressions ( [ wavetn ] ) and ( [ wavern ] ) respectively .
if the signal impinging on the system from the left is represented by an @xmath81-component vector @xmath82 , the output vector @xmath83 simply becomes , for the case of transmission @xmath84 since in this model the matrices @xmath73 and @xmath85 have exactly the same symmetries , they will have the same eigenvectors ; those will also be the eignevectors of the complicated matrices @xmath79 and @xmath86 , since they are expressed ( through the loop expansion ) as products of @xmath87 and @xmath74 .
the elements of the matrix @xmath79 involved in ( [ osubj ] ) can be evaluated in terms of those eigenvectors and the eigenvalues of @xmath73 and @xmath74 through standard algebraic techniques .
the problem is , then , essentially reduced to solving the eigenvalue problem of an @xmath72 matrix . to conclude we will illustrate the treatment of the two dimensional problem by considering the particularly simple case of periodic boundary conditions in the lateral sides of the system ( rows 1 and @xmath71 ) .
we will calculate the total transmission through a strip of length @xmath16 .
then , @xmath88 where the @xmath89 stand for the eigenvalues of the matrix @xmath79 , and the @xmath90 for the corresponding eigenvectors of the matrices @xmath73 and @xmath74 . from the symmetry of the problem readily follows that one of the eigenvectors is @xmath91 and all the others cancel when summed over @xmath77 in ( [ formtn ] ) by orthogonality , leading to @xmath92 .
the eigenvalue @xmath93can be simply obtained by replacing in the expansion ( [ wavetn ] ) or its equivalent for a prw the coefficients @xmath0 and @xmath1 by the eigenvalues @xmath94and @xmath95(associated with @xmath96 ) @xmath97of the matrices @xmath73 and @xmath74 .
those eigenvalues are simply @xmath98 for any row @xmath76 , and a similar expression for @xmath74 .
for the case of a prw@xcite , ( where the walker has a probability @xmath0 of moving forward , a probability @xmath1 of moving back , and a probability @xmath99 of turning to either side , with @xmath100 ) , these eigenvalues can be obtained writing all the possible paths starting from position @xmath76 that lead to transmission ( reflection ) through _ any _ row of a 1-column system : @xmath101 the expansion in this expression corresponds to all paths with no reflection in the vertical direction , one reflection in the vertical direction , etc . ) .
this is simply @xmath102 .
analogously we obtain @xmath103 .
the total transmission for a system of length n becomes , then : @xmath104 figure 5 illustrates the transmission through a strip of length 8 , for periodic , absorbing and reflecting boundary conditions , and for a 1-d chain of the same length , in the specific case @xmath105 .
the inset illustrates the effect of snaking , the existence of a non - zero total transmission even in the case of a zero forward transmission @xmath0 at the level of an individual scattering .
charles wizenfeld assisted with the numerical simulations .
i benefited from helpful discussions with juanjo saenz .
the hospitality of the departamento de fisica de la materia condensada , where the last part of this work was written is gratefully acknowledged . partially supported by the summer faculty fellowship of the office of the vice president of academic affairs of yeshiva university . | it is possible to discuss the propagation of an electronic current through certain layered nanostructures modeling them as a collection of random one - dimensional interfaces , through which a coherent signal can be transmitted or reflected while being scattered at each interface .
we present a simple model in which a persistent random walk ( the t - r model in 1-d ) is used as a representation of the propagation of a signal in a medium with such random interfaces . in this model all the possible paths through the system leading to transmission or reflection
can be enumerated in an expansion in the number of loops described by the path .
this expansion allows us to conduct a statistical analysis of the length of the paths for different geometries and boundary conditions and understand their scaling with the size of the system . by tuning the parameters of the model it is possible to interpolate smoothly between the ballistic and the diffusive regimes of propagation .
an extension of this model to higher dimensions is presented .
we show monte carlo simulations that support the theoretical results obtained . |
an empirical description of the milky way s present structure and formation history requires accurate and consistent age estimates for large samples of stars distributed throughout the galaxy .
although we have recently entered an era of extensive spatial , kinematic , and chemical information beyond the solar neighborhood , comparably extensive age constraints remain elusive .
stellar age is a property that must be inferred from observations with the help of stellar evolution models ; generally , it can not be measured directly " .
therefore , results are inherently limited by the applicability and accuracy of the model used ( see @xcite for a comprehensive review . ) as stellar ages are difficult to measure directly , abundances such as and are commonly used as an age - dating proxy ( e.g. via making maps of mono - age populations ; see @xcite and @xcite ) because the determination of photospheric abundances from spectra is more straightforward . unfortunately for milky way studies , the population of stars that is most readily observable throughout the galaxy red giant stars
is also the one for which it is particularly challenging to estimate ages .
the standard technique of isochrone fitting is impractical in this regime , because there is too much uncertainty both in stellar parameter measurements and in the model isochrones . in other words , stars with very different ages
can have very similar atmospheric parameters and luminosities @xcite . instead , the key to age - dating red giant stars is mass . because the red giant phase is so short ,
the age of a star is essentially its main sequence lifetime , which is set by the star s mass and metallicity @xcite . given mass and metallicity
, one can estimate age using isochrones , e.g. by interpolating between them .
recently , asteroseismology has made it possible to measure masses for giants out to large distances .
the _ kepler _ mission @xcite has conducted long - cadence photometry for over ten thousand giants along a pencil - beam through the galaxy @xcite . from detailed light curves
one can measure two characteristic variability frequencies that directly probe the ( age - dependent ) structure of the stellar interior : @xmath6 is the frequency corresponding to the maximum oscillation power , and @xmath7 is the frequency spacing between two consecutive modes of the same spherical degree .
this approach is especially effective for giants because they have higher densities and thus a larger sound speed , which makes these ( acoustic ) oscillations more pronounced @xcite .
together with a measurement of the star s , and the solar values @xmath8 and @xmath9 , the mass of the star can be estimated using seismic scaling relations .
note that these scaling relations are based on sun - like stars , and may not be suitable for metal - poor stars @xcite .
furthermore , the population of stars with asteroseismic measurements is spatially limited .
@xcite and @xcite greatly expanded the spatial coverage of giants with age estimates by determining masses spectroscopically : they showed that the masses ( and implied ages ) of post dredge - up giants can be measured from high - resolution infrared ( apogee , r@xmath1022,500 ) spectra , and determined a model of mass and age as a function of , , , , and values ( see tables a2 and a3 in @xcite ) .
their work increased the sample of giant stars with known ages to 70,000 , the largest ( and most spatially extended ) sample of stellar ages to date . in this work , we set out to extend this spectroscopic mass & age work to lamost , the largest ongoing stellar spectroscopic survey .
lamost represents a large expansion over apogee in area coverage ( lamost stars are measured away from the disk , unlike apogee ) , sample size , and parameter range ( in particular , ) .
@xcite have shown that basic parameters ( , , , and ) consistent with apogee values can be determined directly from lamost spectra , using _
the cannon_. _ the cannon _ @xcite is a data - driven method for measuring stellar labels " ( a term that refers collectively to all attributes of a star , e.g. physical parameters and element abundances ) from stellar spectra in the context of large spectroscopic surveys .
it has shown promise for bringing qualitatively different stellar surveys onto a consistent physical scale , and for transferring label systems from one survey to another . in @xcite , for example , _ the cannon _ was used to transfer labels from a high - resolution , high- survey ( apogee ) to a low - resolution , modest- survey ( lamost ) , enabling the measurement of the first - ever values from lamost spectra , and the largest and most spatially - extended sample of values to date ( @xmath11450,000 giants ) .
of course , this data - driven approach can only be applied to a subset of the lamost dataset , restricted by the overlap in label space with apogee : because the set of objects in common between apogee and lamost is entirely giants , our model is limited in its applicability to giants , which is only 20% of the lamost dr2 dataset . _
the cannon _ operates in two s.png : a training step and a test step . for a complete description of the methodology ,
see @xcite ; we summarize briefly here . in the training step , _ the cannon _ uses a reference set of objects observed in common between the surveys of interest to construct a predictive model of spectra independently at each wavelength .
for example , for a set of objects measured in common between survey a and survey b , _ the cannon _ might fit a model that predicts every pixel in a survey a spectrum given survey b labels . in the test step , this model can be used to infer _ new _ labels directly from survey a spectra that are by construction on the survey b label scale . _ the cannon _
uses no explicit physical stellar models , is very fast , and achieves comparable accuracy to existing survey pipelines using significantly lower snr spectra ; it requires only a set of objects observed in common between the surveys .
taken together , the work in @xcite , @xcite , and @xcite raises the prospect that ages could be determined for a large sample of lamost giants . in theory
, it seems plausible that mass ( and implied age ) information could be encoded in optical spectra .
after all , in the near - ir , mass is encoded in cn and co molecular regions ; as and features are prominent in the blue parts ( @xmath12 ) of giant spectra ( e.g. @xcite ) it seems plausible that this information could be encoded in lamost spectra too .
recent theoretical work by @xcite and @xcite lends physical justification to why these features should be indicative of mass : during a star s main sequence lifetime , the cno cycle in its core determines the final relative abundances of carbon and nitrogen . after arriving on the giant branch , the material in the core
is dredged up to the surface via convective mixing .
the depth of the convective envelope , and the ratio in the core , is determined by the mass of the star .
thus , in giants that have undergone dredge - up once ( that is , they have not undergone additional convective mixing ) the ratio observed in the photosphere is ( together with metallicity ) highly predictive of mass
. however , and have not previously been measured from low - resolution resolution ( @xmath13 ) spectra . in this work , we extend the apogee - lamost label transfer work of @xcite by two additional labels ( and ) to learn about the information content of lamost spectra .
we use the theoretical relations in @xcite to determine masses and ages for as many giant stars as possible , restricted primarily by the parameter regime in which the relations are applicable .
this will enable us to measure the largest sample of stellar ages to date .
our procedure for transferring labels from apogeeto lamost using _ the cannon _ closely resembles the work of @xcite . here , too , the data consist of spectra from lamost , which we again normalize in a consistent way using a running gaussian , and labels from apogee dr12 . as before ,
the spectral model is quadratic in the labels .
there are , however , important differences in ( and new components to ) our modeling : in the labels that we use , in the reference objects that we use to train the model , in spectral regions that we mask out , and in the incorporation of photometry . here ,
our model is quadratic in seven labels instead of the original five labels : we use , , , , , , and k - band extinction . because we will eventually use the relations in @xcite to translate our carbon and nitrogen abundances into age estimates , our labels need to be on the same scale as those that were used to fit for the relations
thus , whereas we used the calibrated dr12 parameters in @xcite ( those in the ` param ` array ) in this case we use the raw , uncalibrated values from the ` fparam ` array .
furthermore , we do not use the full reference set of 9956 objects from @xcite , because some of these have unreliable and reference labels .
following @xcite , we excise objects that have both @xmath14 and @xmath15 $ } } < -0.5 $ ] ( 743 objects ) in order to eliminate objects with only an upper limit measurement on and a lower limit on . in addition , @xcite found that the minimum possible to measure is on the level of -0.4 to -0.5 dex , so we also exclude objects with @xmath16 $ } } < -0.4 $ ] ( 40 objects ) .
this left us with 9173 out of the original 9956 . in @xcite , we fit an independent spectral model at _ every _ spectral pixel . however , there are features in lamost spectra that arise from effects in a different velocity system from that of the star : for example , diffuse interstellar bands ( dibs ) and the na i doublet are interstellar absorption features originating from intervening material .
we noticed by examining the leading coefficients of an initial cannon model ( see figure [ fig : leading - coeffs ] ) that _ the cannon _ was `` learning '' to use these features to predict labels intrinsic to the star , particularly , and that this introduced radial velocity - dependent systematic errors into the label estimates - enhanced stars have a different line - of - sight velocity distribution , velocity and may well be correlated ] .
the leading coefficients of an initial _
the cannon _ model also indicated that the imperfectly corrected telluric bands , originating in the earth s atmosphere , left small , but significant , velocity - dependent effects in the rest - frame spectra of the stars . to prevent _ the cannon _ from using these features spuriously , we masked them out by setting the inverse variances corresponding to these pixels to be zero . to be conservative ,
roughly half of each spectrum was masked , but because the most important features for our labels of interest were preserved , this still improved our label estimates . masking out the interstellar absorption features took out most of the spectral information on , which originated in the dib strength .
but of course , many - band photometry of a star encodes a combination of its effective temperature and its reddening .
we found that incorporating photometry not only enabled us to accurately and precisely predict for our reference objects , but also to improve the precision of our estimate of , particularly for lower- spectra .
it also improved the precision of our estimate of , presumably because measurements of and are highly covariant for spectra of this resolution due to blending ( see e.g. ting et al .
2016b ; submitted to ) .
the scatter in decreased by 70% in the lowest- spectra , by 50% in spectra with @xmath17 , and by 25% in spectra with @xmath18 . for low- spectra , the scatter in and improved by up to 20% , although for @xmath19 the difference was negligible .
we incorporated photometry as follows : we took magnitudes ( and the associated uncertainties ) from eight bands , taken from apass dr9 @xcite , 2mass @xcite and wise @xcite : _ g , r , i , j , h , k , w1 _
( 3.4@xmath20 m ) , _ w2 _
( 4.6@xmath20 m ) . from these
, we constructed seven colors : _ g - r _ , _ r - i _ , _ i - j _ , _ j - h _ , _ h - k _ , _ k - w1 _ , and _ w2-w1_. for each reference object , we added its seven colors as pixels " to its spectrum : the color as the flux " and the uncertainty as the error bar . "
using colors restricted us to the set of reference objects with apass , 2mass & wise magnitudes : 8472 of the 9173 .
we emphasize that _
the cannon _ builds a model of spectra but is agnostic to whether the value of a `` spectral pixel '' is a true flux value or simply another observed property of the star that is sensitive to the labels of interest .
note , however , that our quadratic model is probably not sufficiently complex for the photometric data ; we would expect improvements for a model that is more complex at the photometric pixels relative to the spectroscopic pixels . as in @xcite
, we evaluate the accuracy and precision of our model using a leave-@xmath21-out " cross - validation test .
we split the 8472 reference objects into eight groups , by assigning each one a random integer between 0 and 7 .
we leave out each group in turn , and train a model on the remaining seven groups .
we then apply that model to infer new labels for the group that was left out .
we use the results of the cross - validation to determine which objects are appropriate reference objects for training the model : we excise objects whose difference from the training value is greater than four times the scatter in that label , leaving 8125 objects .
we train the model on these 8125 objects and re - run the cross - validation .
we also use the model to infer labels for the 347 objects excised from the training , in order to properly account for all of the objects in the following error analysis . at the end of this process ,
each of the 8472 objects has a new set of labels determined by _
the cannon _ , from a model that was _ not _ trained using that object .
figure [ fig : cross - validation ] shows the comparison of these cannon - inferred `` test '' labels with the reference labels used in training , for high objects ; there is a significant decrease in scatter compared to the corresponding figure in @xcite ( fig .
6 ) . for objects with @xmath18 ,
the labels are consistent with the apogee training values to within 53k in , 0.11dex in , 0.05dex in , 0.06dex in , 0.09dex in , 0.03dex in , and 0.04mag in .
these are comparable to , or within , the stated aspcap uncertainties @xcite .
figure [ fig : snr - test ] shows the scatter in different bins of ( where is the median value of formal across all pixels in the spectrum ) in all of the labels except for ( which is primarily determined from the additional photometric pixels , not taken into account in determining the ) . by construction , as a result of this data - driven label transfer , there is significant improvement in agreement with the apogee values over those from the lamost pipeline . .
the low bias and scatter in and demonstrates that these abundances can in fact be measured from low resolution lamost spectra . ] , not the reported lamost error bar . ] a key strength of _ the cannon _ is the physical interpretability of the spectral model .
an independent model is fit at every pixel of the spectrum , so each pixel has a set of model coefficients as well as a model scatter term .
the leading ( linear ) coefficient in each label can be thought of as a proxy for how sensitive a particular spectral pixel is to that particular label ; thus , each label has a wavelength - dependent indicator of which spectral regions are most informative .
each pixel also has a model scatter term ; this is the intrinsic variance in the model at that pixel , as distinct from the observational variance . in other words , it is the expected deviation from the model at that particular pixel , in the limit of a perfect measurement .
these are shown in figure [ fig : leading - coeffs ] , in the part of the spectrum found by _
the cannon _ to be most predictive of labels ( the blue end , @xmath22 ) , together with the scatter in the model .
these linear coefficients are the first derivative of the model at a set of fiducial stellar parameters ; in this case , we pivot the model around the mean value in each label across the training set : = 4687k , = 2.83dex , = -0.19dex , = -0.01dex , = 0.10dex , = 0.09dex , = 0.09mag . to facilitate comparison ,
each derivative has been scaled by the approximate error in the corresponding label : @xmath23k , @xmath24dex , @xmath25$}}\sim 0.05$]dex , @xmath26$}}\sim 0.05$]dex , @xmath27$}}\sim 0.04$]dex , @xmath28$}}\sim 0.07$]dex @xcite .
is the variance intrinsic to the model ; it quantifies each pixel s expected deviations from the model in the limit of a perfect measurement .
we display the model coefficients and scatter blueward of 5800 because this was found by _
the cannon _ to be the region with the most sensitive features . ]
the new labels in this work are and , so we should demonstrate that _ the cannon_measures these values from physically plausible spectral signatures .
ting et al .
2016b ( submitted to ) quantified how the precision with which various abundances ( including and ) can be measured varies as a function of survey resolution .
they showed that low - resolution ( @xmath29 ) spectra , such as those from lamost , have the same theoretically achievable uncertainties per stellar label as medium - resolution ( @xmath30 ) spectra , under the following assumptions : equal exposure time ( so that a low - resolution spectrum has higher per resolution element ) , an equal number of detector pixels ( so that low - resolution spectra have more extensive wavelength coverage ) , and a constant sampling per resolution element . these predictions are based in part on gradient spectra calculated using kurucz models @xcite , which are assumed to be perfect .
of course , this aspect does not pertain to data - driven models ( see the discussion in @xcite ) . to make a direct comparison between the cannon model and theoretical predictions
, we calculate gradient spectra and compare them to the gradient spectra calculated from kurucz models by @xcite .
gradient spectra are a quantification of how much the flux at a given wavelength changes with changes to a given label : in other words , it characterizes the sensitivity or information content of each wavelength for a given label .
following equation 2 in ting et al .
2016b , gradient spectra are calculated as follows : @xmath31 where @xmath32 represents a model spectrum across wavelengths @xmath33 , generated using a set of labels @xmath34 . a fractional change in a particular label @xmath35 results in a fractional change in the spectrum @xmath36 at each wavelength @xmath33 .
in other words , to study sensitivity of a spectral region to a particular label , one changes the value in that label and calculates the fractional change in the flux in that region .
we have two sets of model spectra : one from the cannon model as described in section [ sec : procedure ] , and one from the kurucz models , in both cases generated using labels for a solar metallicity k - giant ( = 4800 , = 3.5 ) .
we use each of these models to calculate a gradient spectrum for by varying by 0.2dex , and a gradient spectrum for by varying by 0.2dex . for a more direct comparison ,
we normalize the theoretical gradient spectra the same way as the lamost spectra . figure [ fig : cn - gradient ] shows the cannon model gradient spectra overlaid with the kurucz model gradient spectra from ting et al . 2016b , for the cn and ch molecular features in the wavelength range 4100 - 4400 ( see @xcite ) .
the panel on the left shows the gradient spectra for carbon , and the panel on the right shows the gradient spectra for nitrogen . as the theoretical gradient spectra ( red ) were generated purely from physical models , and the cannon gradient spectra ( black ) represent an entirely data - driven relationship between these wavelength regions and abundances from apogee , the qualitative similarity between them is gratifying .
there are clearly some quantitative differences , but we simply seek to demonstrate here that the signatures of carbon and nitrogen from the data - driven cannon model come from astrophysically sensible wavelength regions , such as the 4215 cn band .
furthermore , the differences between the two panels demonstrates that we are measuring each element from distinct features , not simply correlations between the two ( e.g. the carbon - sensitive ch ( g ) band , not present in the nitrogen signature ) . the fact that and
share regions of sensitivity does not mean that they are degenerate ; they may be covariant , but can still be independently measured when fit for simultaneously ( see ting et al . 2016b , submitted ) .
the qualitative similarity between the black and red lines demonstrates that the cannon measurements of and are coming from astrophysically sensible spectral regions , e.g. the 4215 cn band .
furthermore , the difference between the left and right panels demonstrates that and are being measured independently , not just via correlations : for example , the nitrogen signature does not include any changes in the g ( ch ) band . ]
to transform and to mass and age , we use the formulas characterized by the coefficients in table a2 and table a3 of @xcite , which are in turn based on asteroseismic mass measurements for stars with and measurements .
these relations are only applicable within a certain range of label values , restricting the number of objects for which we can infer masses and ages via their and .
although we infer and for the full set of 454,180 test objects described in @xcite , we apply the following cuts ( following @xcite ) which leave 230,901 objects suitable for applying the formula .
this sets the primary restriction on the size of our mass and age catalog ; for example , although the lamost data contain a large population of low - metallicity outer disk stars , we can not estimate masses and ages for those objects . @xmath37 $ } } < 0.25 \\ 4000 < { \mbox{$\rm t_{eff}$ } } < 5000 \\ 1.8
< { \mbox{$\rm \log g$ } } < 3.3 \\
-0.25 < { \mbox{$\rm [ c / m]$ } } < 0.15 \\
-0.1 < { \mbox{$\rm [ n / m]$ } } < 0.45 \\
-0.05 < { \mbox{$\rm [ \alpha / m]$ } } < 0.3 \\ -0.1 < { \mbox{$\rm [ ( c+n)/m]$ } } < 0.15 \\
-0.6 < { \mbox{$\rm [ c / n]$ } } < 0.2 \\ \end{cases}\ ] ] the estimated masses and ages have uncertainties that come from the intrinsic scatter of the relation @xcite and the individual stellar label uncertainties . to estimate the latter , we sample from each star s label _ pdf _ 100 times , approximating each distribution as a gaussian with a standard deviation equal to the scatter in that label . that scatter is a function of signal to noise ( see figure [ fig : snr - test ] ) ,
so for each object we take the standard deviation of its gaussian spread to be the scatter associated with that signal to noise bin .
thus , each object has a distribution in mass and age .
we take the mass and age value to be the median of that distribution , and estimate the uncertainty using the half - width of the central 68@xmath38 percentile .
this procedure does not account for the errors in the training labels nor the scatter in the @xcite relation .
there are , however , additional systematic errors from the age relation in @xcite that are not taken into account here .
note that these are all distinct from , an in addition to , the formal error from the cannon model fit .
we provide a catalog of all our inferred labels , including mass and age ; an excerpt is shown in table [ tab : results ] .
we provide and for all 454,364 objects , but mass and age only for the 230,901 of those that fall within the label space of @xcite .
we also provide flags indicating whether an object was used as a reference object and whether the object falls within the label range of @xcite .
we also provide the formal for the spectrum of each object and the reduced chi squared of the fit , which is the chi squared divided by the approximate number of pixels in each spectrum ( @xmath39 ) .
note that the and the reduced chi squared are both low by roughly a factor of three ; see the discussion in section 4.1 of @xcite .
furthermore , note that the values of , , , and will not be identical to their corresponding values in @xcite for several reasons : they are on the uncalibrated apogee label scale , and there have been various changes in our procedure ( masking 50% of the spectrum , the inclusion of photometry , fitting for additional labels ) .
cccccccccccc lamost i d & ra & dec & & & & & & & & mass & log(age ) + & ( deg ) & ( deg ) & ( k ) & ( dex ) & ( dex ) & ( dex ) & ( dex ) & ( dex ) & mag & ( @xmath40 ) & dex + spec-55859-f5902_sp01 - 034 & 331.92 & -1.78 & 4794 & 3.22 & -0.507 & 0.0645 & -0.0242 & 0.228 & 0.0540 & 0.78 & 1.0 + spec-55859-f5902_sp03 - 209 & 331.14 & 0.853 & 4620 & 2.88 & -0.347 & 0.0984 & 0.107 & 0.220 & 0.0131 & 1.0 & 0.85 + spec-55859-f5902_sp06 - 160 & 334.27 & -0.159 & 4240 & 2.23 & -0.293 & 0.0734 & 0.102 & 0.208 & 0.148 & 1.3 & 0.66 + spec-55859-f5902_sp08 - 146 & 333.41 & -0.397 & 4895 & 3.29 & -0.337 & -0.0221 & -0.0243 & 0.212 & 0.0293 & 1.2 & 0.65 + cccccccccc lamost i d & @xmath41 ( ) & @xmath41 ( ) & @xmath41 ( ) & @xmath41 ( ) & @xmath41 ( ) & @xmath41 ( ) & @xmath41 ( ) & @xmath41(mass ) & @xmath41(log(age ) ) + & ( k ) & ( dex ) & ( dex ) & ( dex ) & ( dex ) & ( dex ) & ( mag ) & ( @xmath40 ) & ( dex ) + spec-55859-f5902_sp01 - 034 & 3290 & 0.010 & 0.0027 & 0.0043 & 0.0089 & 0.00066 & 0.00036 & 0.33 & 0.34 + spec-55859-f5902_sp03 - 209 & 73.8 & 0.00031 & 8.0e-5 & 0.00012 & 0.0003 & 2.67e-5 & 5.2e-5 & 0.097 & 0.12 + spec-55859-f5902_sp06 - 160 & 65.0 & 0.0004 & 0.0001 & 7.9e-5 & 0.00014 & 4.93e-5 & 7.0e-5 & 0.43 & 0.34 + spec-55859-f5902_sp08 - 146 & 5150 & 0.016 & 0.0047 & 0.0042 & 0.0058 & 0.0015 & 0.00041 & 0.47 & 0.45 + ccc lamost i d & snr & red . + & & @xmath42 + spec-55859-f5902_sp01 - 034 & 33.7 & 0.44 + spec-55859-f5902_sp03 - 209 & 169 & 1.7 + spec-55859-f5902_sp06 - 160 & 130 & 1.2 + spec-55859-f5902_sp08 - 146 & 19.9 & 0.51 + we now investigate whether our inferred age values seem astrophysically plausible .
figure [ fig : mh - am - cage ] shows the ( , ) plane color - coded by age for @xmath43 objects with @xmath44 .
shown is the mean age in each bin , weighted by the estimated uncertainty in the age measurement , for bins with a minimum of 20 objects .
we see an astrophysically sensible age gradient with changing abundances , from the young , low- sequence to the old , high- sequence .
this is qualitatively very similar to the gradient seen from small high - resolution datasets of main sequence turn - off stars in the solar neighborhood ( e.g. @xcite ) .
furthermore , as figure [ fig : mass - age - comparison ] shows , our masses and ages ( for the reference objects ) are in remarkable agreement with the masses and ages from the @xcite catalog , determined via a rather different approach . in that approach ( x axis )
masses were measured directly from apogee spectra ( @xmath45 ) and ages were estimated via isochrone fitting . in our approach ( y axis )
we measured and directly from lamost spectra ( @xmath46 ) and used the relations in @xcite to estimate ages .
this agreement supports the plausibility of our estimates .
finally , figure [ fig : full - ages ] shows the enhanced spatial distribution of our sample over that from 70,000 stars in apogee @xcite .
as expected , younger stars are concentrated towards the disk mid - plane , and older stars extend to a larger scale height away from the disk and into the bulge and halo .
( leaving 42420 objects ) and only show bins with @xmath47 objects .
the gradient from young @xmath48-poor to old @xmath48-rich stars is astrophysically very plausible . ]
70,000 objects ) and the bottom panel overlays these values with 230,901 ages inferred via and by _ the cannon _ from the lamost spectra . the much more extensive area coverage of the lamost data is immediately apparent .
using a data - driven approach to spectral modeling , and fitting for all labels simultaneously , we find that we can measure accurate and precise carbon and nitrogen abundances from low - resolution ( @xmath2 ) lamost spectra . for post dredge - up giants , as in the sample from @xcite , these and measurements enable mass and age estimates across the sky , to 0.08dex in mass and to 0.2dex in age . with this new set of ages
, we have a very different spatial sampling than apogee : we have essentially tied in - the - disk and off - the - disk ages onto the same scale , as lamost has a much better sampling of the thick disk than apogee .
the success of our data - driven approach in extracting information on and from blended regions ( see figure [ fig : cn - gradient ] ) holds promise for a natural extension of this work to measuring additional individual element abundances . _
the cannon _ has already been successful at measuring individual abundances from apogee spectra , in part because the model is not restricted to unblended element windows @xcite . indeed , ting et al .
2016b predicted using theoretical models that spectra of comparable resolution to lamost should not only contain sufficient information to precisely constrain and , but also a large suite of other individual element abundances , such as aluminum , calcium , manganese , and nickel .
for the purpose of and measurement in this work , it was helpful to apply broad masks to the spectra , to fully remove telluric and interstellar absorption features . depending on which spectral regions encode information on [ x / h ] , however , fitting for these additional labels would likely require more precise masking in order to avoid removing important signatures .
the quality of the data reduction may limit which individual [ x / h ] can be returned . finally , at nearly identical values of \ { , , } on the giant branch , mass , or age , is highly predictive of luminosity .
age constraints to 0.2dex could therefore be useful for improving estimates of stellar luminosity , and thus distance .
we are grateful to evan kirby ( caltech ) for his careful reading and constructive feedback on the manuscript .
it is furthermore a pleasure to thank jon bird ( vanderbilt ) , sven buder ( mpia ) , andy casey ( ioa cambridge ) , christina eilers ( mpia ) , lynne hillenbrand ( caltech ) , sarah martell ( unsw ) , and marie martig ( mpia ) for helpful discussions and assistance .
ayqh is grateful to the community at the mpia for their support and hospitality during the period in which much of this work was performed .
ayqh was supported by a national science foundation graduate research fellowship under grant no .
mkn and hwr have received funding for this research from the european research council under the european union s seventh framework programme ( fp 7 ) erc grant agreement n. [ 321035 ] .
dwh was partially supported by the nsf ( grant iis-1124794 ) , nasa ( grant nnx08aj48 g ) , and the moore - sloan data science environment at nyu .
cl acknowledges the strategic priority research program
the emergence of cosmological structures " of the chinese academy of sciences , grant no .
xdb09000000 , the national key basic research program of china 2014cb845700 , and the national natural science foundation of china ( nsfc ) grants no .
11373032 and 11333003 . guoshoujing telescope ( the large sky area multi - object fiber spectroscopic telescope lamost ) is a national major scientific project built by the chinese academy of sciences .
funding for the project has been provided by the national development and reform commission .
lamost is operated and managed by the national astronomical observatories , chinese academy of sciences .
funding for the sloan digital sky survey iv has been provided by the alfred p. sloan foundation , the u.s .
department of energy office of science , and the participating institutions .
sdss - iv acknowledges support and resources from the center for high - performance computing at the university of utah .
the sdss web site is www.sdss.org .
sdss - iv is managed by the astrophysical research consortium for the participating institutions of the sdss collaboration including the brazilian participation group , the carnegie institution for science , carnegie mellon university , the chilean participation group , the french participation group , harvard - smithsonian center for astrophysics , instituto de astrofsica de canarias , the johns hopkins university , kavli institute for the physics and mathematics of the universe ( ipmu ) / university of tokyo , lawrence berkeley national laboratory , leibniz institut fr astrophysik potsdam ( aip ) , max - planck - institut fr astronomie ( mpia heidelberg ) , max - planck - institut fr astrophysik ( mpa garching ) , max - planck - institut fr extraterrestrische physik ( mpe ) , national astronomical observatory of china , new mexico state university , new york university , university of notre dame , observatrio nacional / mcti , the ohio state university , pennsylvania state university , shanghai astronomical observatory , united kingdom participation group , universidad nacional autnoma de mxico , university of arizona , university of colorado boulder , university of oxford , university of portsmouth , university of utah , university of virginia , university of washington , university of wisconsin , vanderbilt university , and yale university .
this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . | we measure carbon and nitrogen abundances to @xmath0dex for @xmath1 giant stars from their low - resolution ( @xmath2 ) lamost dr2 survey spectra .
we use these and measurements , together with empirical relations based on the apokasc sample , to infer stellar masses and implied ages for 230,000 of these objects to @xmath3dex and @xmath4dex respectively .
we use _ the cannon _ , a data - driven approach to spectral modeling , to construct a predictive model for lamost spectra .
our reference set comprises 8125 stars observed in common between the apogee and lamost surveys , taking seven apogee dr12 labels ( parameters ) as ground truth : , , , , , , and .
we add seven colors to the cannon model , based on the _
g , r , i , j , h , k , w1 , w2 _ magnitudes from apass , 2mass & wise , which improves our constraints on and by up to 20% and on by up to 70% .
cross - validation of the model demonstrates that , for high- objects , our inferred labels agree with the apogee values to within 50k in temperature , 0.04 magnitudes in , and @xmath5dex in , , , , and .
we apply the model to 450,000 giants in lamost dr2 that have _ not _ been observed by apogee .
this demonstrates that precise individual abundances can be measured from low - resolution spectra , and represents the largest catalog of , , masses and ages to date .
as as result , we greatly increase the number and sky coverage of stars with mass and age estimates . |
magnetic reconnection is the physical process of breaking and rearrangement of magnetic field lines , which changes the topology of the field .
it is one of the most fundamental processes of plasma physics and is believed to be at the core of many dynamic phenomena in laboratory experiments and in cosmic space .
unfortunately , in spite of being so important , magnetic reconnection is still relatively poorly understood from the theoretical point of view .
the reason is that plasmas usually have very high temperatures and low densities . in such plasmas ,
the spitzer resistivity is extremely small and magnetic fields are almost perfectly frozen into cosmic plasmas . as a result , simple theoretical models , such as the sweet - parker reconnection model @xcite predict that the magnetic reconnection processes should be extremely slow and insignificant throughout the universe . on the other hand ,
astrophysical observations show magnetic reconnection tends to be fast and is likely to be the primary driver of many highly energetic cosmic processes , such as solar flares and geomagnetic storms .
this contradiction between theoretical estimates and astrophysical observations triggered multiple attempts to build a theoretical model of fast magnetic reconnection .
first , in 1964 petschek proposed a fast reconnection model @xcite , in which fast reconnection is achieved by introducing switch - off magnetohydrodynamic ( mhd ) shocks attached to the ends of the reconnection layer in the downstream regions and by choosing the reconnection layer length to be equal to its minimal possible value under the condition of no significant disruption to the plasma flow .
however , later numerical simulations and theoretical derivations did not confirm the petschek theoretical picture for the geometry of a reconnection layer @xcite .
second , numerical simulations studies of anomalous magnetic reconnection , for which resistivity is enhanced locally in the reconnection layer , were pioneered by ugai and tsuda @xcite , by hayashi and sato @xcite , and by scholer @xcite .
third , lazarian and vishniac proposed that fast reconnection can occur in turbulent plasmas @xcite , although the back - reaction of magnetic fields can slow down reconnection in this case @xcite . finally , recently there have been number of attempts to explain the fast magnetic reconnection by considering non - mhd effects @xcite .
most previous theoretical and numerical studies concentrated on reconnection processes in two - dimensions or in `` two - and - a - half - dimensions '' .
the later is the term used for a problem in which physical scalars and all three components of physical vectors depend only on two spatial coordinates ( e.g. @xmath0 and @xmath1 ) and are independent of the third coordinate ( @xmath2 ) . in this paper
we consider two - and - a - half - dimensional magnetic reconnection with anomalous resistivity in the classical sweet - parker - petschek reconnection layer , which is shown in the left plot in fig .
[ fig_reconnection_layer ] .
the reconnection layer is in the x - y plane with the y - axis being along the layer and the x - axis being perpendicular to the layer .
the length of the layer is equal to @xmath3 .
note that @xmath4 is approximately equal to or smaller than the global magnetic field scale , which we denote as @xmath5 .
the thickness of the classical reconnection layer , @xmath6 , is much smaller than its length , i.e. @xmath7 .
the classical sweet - parker - petschek reconnection layer is assumed to possess a point symmetry with respect to its geometric center point o and reflection symmetries with respect to the axes @xmath0 and @xmath1 ( refer to fig . [ fig_reconnection_layer ] ) .
thus , for example , the x- and y - components of the plasma velocities @xmath8 and of the magnetic field @xmath9 have the following simple symmetries : @xmath10 , @xmath11 , @xmath12 and @xmath13 .
there might be a pair of petschek shocks attached to each of the two reconnection layer ends in the downstream regions ( see fig .
[ fig_reconnection_layer ] ) .
because of the mhd jump conditions on the petschek shocks , the presence of these shocks requires the presence of a significant perpendicular magnetic field @xmath14 at the reconnection layer ends @xcite .
the plasma outflow velocity from the reconnection layer is approximately equal to the alfven velocity @xmath15 ( if the plasma viscosity is not very large ) . the plasma inflow velocity @xmath16 outside of the reconnection layer , at point m in fig
. [ fig_reconnection_layer ] , is much smaller than the outflow velocity , @xmath17 .
finally , the magnetic field outside the reconnection layer is mostly in the direction of the layer ( i.e. in the y - axis direction ) .
the problem of quasi - stationary anomalous magnetic reconnection in the classical sweet - parker - petschek reconnection layer was recently theoretically addressed by kulsrud @xcite for the special case of zero guide field ( @xmath18 ) , zero plasma viscosity and anomalous resistivity that is a piecewise linear function of the electric current [ see eq .
( [ linear_resistivity ] ) ] .
there are two major results of the kulsrud anomalous reconnection model .
first , in the case of the petschek geometry of the reconnection layer and a constant resistivity , one has to calculate the half - length of the reconnection layer @xmath4 from the mhd equations and the jump conditions on the petschek shocks , instead of treating @xmath4 as a free parameter ( as petschek did ) .
when the layer half - length @xmath4 is calculated correctly , it turns out to be approximately equal to the global magnetic field scale , @xmath19 . in this case
the petschek reconnection reduces to the slow sweet - parker reconnection @xcite .
this theoretical result is in agreement with the results of numerical simulations of two - dimensional reconnection with constant resistivity @xcite .
the second major result of the kulsrud reconnection model is that in the case when resistivity is anomalous and enhanced ( e.g. by plasma instabilities ) , the reconnection rate becomes considerably faster than the sweet - parker rate . in this paper
we develop and use a new theoretical approach for calculation of the reconnection rate in the case of anomalous resistivity .
this approach is based on application of the mhd equations in a small region , which is localized at the geometric center of a thin reconnection layer ( point o in fig .
[ fig_reconnection_layer ] ) and has a size of order of the layer half - thickness @xmath20 .
it turns out that by using local analytical calculations in a thin reconnection layer , we can derive an accurate and rather precise estimate for the reconnection rate .
in particular , we find the interesting and important result that a quasi - stationary reconnection rate is fully determined by the anomalous resistivity function and by the magnetic field configuration just outside the reconnection layer ( at point m in fig .
[ fig_reconnection_layer ] ) .
the underlying physical foundations of our analytical derivations are , of course , the same as those previously used by others @xcite . however , our theoretical and computational approach is somewhat different from the conventional approach to the reconnection problem , and we explain the difference in the next section .
there are three main benefits of our novel theoretical approach to magnetic reconnection .
first , our approach allows us to extend the results for anomalous reconnection obtained by kulsrud in 2001 @xcite to a general anomalous reconnection case , when the guide field and plasma viscosity are arbitrary and anomalous resistivity @xmath21 is an arbitrary function of the electric current and the two spatial coordinates .
second , it gives a new important insight into the reconnection problem , such as dependence of the reconnection rate on magnetic field configuration just outside the reconnection layer .
third , our approach , based on local calculations , is applicable to cases when there is no well - defined global magnetic field structure , such as the case of multiple current sheets in a turbulent plasma . in our calculations
we make only a few assumptions , which are described in detail in the beginning of sec .
[ new_model ] .
this paper is organized as follows .
because our derivations are rather complicated in the general case of reconnection with anomalous resistivity , we find it useful and instructive to consider the sweet - parker reconnection first and to compare our theoretical approach to the classical sweet - parker calculations in the next section . in sec .
[ new_model ] , we derive our general equations for magnetic reconnection with anomalous resistivity , including the equation for the reconnection rate . in sec .
[ reconnection_limits ] we consider and analyze different special cases of magnetic reconnection , in which our equations simplify and become easier for analysis and comparison to the previous theoretical and simulation results . in sec .
[ simulations ] we present the results of our numerical simulations of unforced anomalous magnetic reconnection .
these simulations are intended for a demonstration of the predictions of our theoretical reconnection model for the special case of the piecewise linear resistivity function that was considered by kulsrud @xcite . finally , in sec . [ conclusions ] we give our conclusions and discuss our results .
some derivations are given in the appendices of the paper .
for simplicity and brevity , hereafter in the paper for all electromagnetic variables we use physical units in which the speed of light and four times @xmath22 are replaced unity , @xmath23 and @xmath24 . to rewrite our equations in the standard cgs units , one needs to make the following substitutions for electromagnetic variables in the equations : magnetic field @xmath25 , electric field @xmath26 , electric current @xmath27 , resistivity @xmath28 ( does not change ) , magnetic field vector potential @xmath29 .
in this section we consider sweet - parker reconnection @xcite .
we assume that resistivity is constant , @xmath30 , the plasma viscosity is zero , the guide field is zero ( @xmath18 ) , and the geometry of the reconnection layer is the classical sweet - parker geometry with the layer half - thickness @xmath31 and the layer half - length @xmath32 , as shown in fig .
[ fig_reconnection_layer ] .
the purpose of this section is to introduce and explain our new theoretical approach to calculation of the reconnection rate and to compare it with the classical sweet - parker calculations . for this purpose
, we first present the classical , conventional derivation of the sweet - parker formula for the reconnection rate , and afterward we present our new derivation of the formula and explain the difference between the two derivations . in the case of a quasi - stationary reconnection the magnetic field in the reconnection region changes slowly in time , @xmath33 .
therefore , in the two - and - a - half - dimensional geometry ( @xmath34 ) the faraday s law equation @xmath35 results in the z - component of the electric field being constant in the reconnection region , @xmath36 and @xmath37 is a function of time only .
on the other hand , in two - and - a - half dimensions the @xmath0- and @xmath1-components of faraday s law equation @xmath38)$ ] reduce to the following equation for the z - component of the magnetic vector potential @xmath39 : -e_z(t)&=&a_z / t= -(*v*)a_z+^2 a_z + & = & v_xb_y - v_yb_x - j_z .
[ mhd_equation ] ( the resistivity @xmath30 is constant in the sweet - parker reconnection case , but equation ( [ mhd_equation ] ) is general and is valid even if resistivity @xmath21 is anomalous and non - constant . ) now , the left- and right - hand - sides of equation ( [ mhd_equation ] ) are constant in space . therefore , the right - hand - side of equation ( [ mhd_equation ] ) is constant across the reconnection layer ( i.e. along the x - axis ) .
we equate its values at points o and m , which are on the x - axis and are shown in fig .
[ fig_reconnection_layer ] . as a result
, we immediately obtain _
oj_o = v_rb_m , [ first_mhd ] where we use the following notations : point o is the geometric center of the reconnection layer , where the z - component of the current is @xmath40 and the plasma velocity is zero .
point m is a point on the x - axis just outside the reconnection layer , where the resistivity term can be neglected in equation ( [ mhd_equation ] ) , see fig .
[ fig_reconnection_layer ] .
we also use the notations @xmath41 and @xmath42 for the y - component of the field and x - component of the plasma velocity at point m , and take the reconnection velocity @xmath16 positive . note that @xmath43 at point m because of the symmetry of the problem with respect to the x - axis .
next , we estimate current @xmath44 at the central point o of the reconnection layer as j_ob_m/_o , [ j_o ] where @xmath20 is the reconnection layer half - thickness , and we use ampere s law , @xmath45 , in which we drop the @xmath46 term because the reconnection layer is thin , then obviously the length of the layer would be of the order of its thickness . ] .
next , consider the x- and y - components of the equation of plasma motion .
we will see below that the sweet - parker reconnection is slow , @xmath47 .
therefore , in the equation of plasma motion along the x - axis ( i.e. across the reconnection layer ) the inertial term can be neglected , and this equation becomes the force balance equation , @xmath48 , resulting in @xmath49 , where @xmath50 and @xmath51 are the values of the plasma pressure at points o and m. here we use the fact that the magnetic field is zero at the central point o because of the symmetry of the problem . as far as the equation of plasma motion along the y - axis ( i.e along the layer ) is concerned , the y - components of the magnetic tension force and the pressure gradient force are approximately equal in the sweet - parker reconnection case .
indeed , on the y - axis the tension force can be estimated as @xmath52 . here
we estimate the x - component of the field as @xmath53 because @xmath14 is produced by the rotation of the @xmath54 field component in the reconnection layer @xcite , and in the sweet - parker model it is assumed that the downstream pressure is equal to the upstream pressure @xmath51 ( see @xcite ) .
the pressure and magnetic tension forces accelerate plasma along the reconnection layer ( i.e. along the y - axis ) up to the downstream velocity @xmath55 , which can be estimated from the energy conservation equation ( 1/2)v_out^2b_m^2/2 v_outv_ab_m/. [ sp_energy ] this equation means that the work done by the pressure and magnetic tension forces along the entire reconnection layer is equal to the kinetic energy of the plasma in the downstream regions .
finally , in the sweet - parker reconnection model the plasma is assumed to be incompressible .
therefore , the mass conservation condition for the entire reconnection layer results in lv_r_ov_out , [ sp_mass ] where @xmath55 , the velocity of plasma outflow in the downstream regions , is given by equation ( [ sp_energy ] ) . using equations ( [ first_mhd])-([sp_mass ] ) , we obtain the formula for the sweet - parker reconnection velocity @xcite , v_rv_a(_o / v_al)^1/2 .
[ sp_speed ] equations ( [ first_mhd])-([sp_speed ] ) are the classical sweet - parker equations .
note that equations ( [ first_mhd ] ) , ( [ j_o ] ) and ( [ sp_speed ] ) are local , in the sense that they are written for a small region of space , which is localized at the geometric center of a thin reconnection layer ( point o in fig .
[ fig_reconnection_layer ] ) and has a size of order of the layer half - thickness @xmath20 .
all physical quantities that enter these three equations are defined in this small region of space . at the same time equations ( [ sp_energy ] ) and ( [ sp_mass ] )
are global , in the sense that they result from consideration of the entire reconnection layer and they include the plasma outflow velocity @xmath56 , which is a physical quantity in the downstream regions at the ends of the reconnection layer . in our new theoretical approach to the calculation of the reconnection rate
we intend to use only local equations .
our intent and our derivations will be justified by the new and important results that we obtain in the next section of this paper and discuss in sec .
[ conclusions ] . at present , let us explain our theoretical approach for the simple case of sweet - parker reconnection that we consider in this section . in the derivation of our theoretical model for magnetic reconnection we keep equations ( [ first_mhd ] ) and ( [ j_o ] ) unchanged because these equations are local . however , we rewrite equations ( [ sp_energy ] ) and ( [ sp_mass ] ) because they are global . to rewrite these two global equations in a local form
, we consider a point @xmath57 that is located on the y - axis in an infinitesimal vicinity of the reconnection layer central point o ( see fig . [ fig_reconnection_layer ] ) and has an infinitesimally small value of its y - coordinate @xmath58 .
since @xmath59 , along the o@xmath60 interval we can use the up - to - the - first - order taylor expansions @xmath61 , @xmath62 and @xmath63 for the values of the perpendicular magnetic field @xmath14 , plasma velocity @xmath64 and z - component of the current @xmath65 .
these expansions are along the y - axis , and , of course , @xmath66 and @xmath67 are the first - order partial derivatives of @xmath14 and @xmath64 at point o ( note that @xmath68 ) .
now equation ( [ sp_energy ] ) for the plasma acceleration along the y - axis can easily be rewritten in a local form at point @xmath60 as ( ) ^2 & & _ 0^yb_x(0,y)j_z(0,y)dy= _ 0^yy j_o dy + & = & j_oy^2/2 ^2j_o , [ sp_local_energy ] where the left - hand - side of this equation is the plasma kinetic energy at point @xmath60 .
on the right - hand - side of equation ( [ sp_local_energy ] ) we keep only the magnetic tension force for plasma acceleration because , as we found above , in the sweet - parker reconnection case the y - components of the magnetic tension and pressure gradient forces are approximately equal to each other ( note that in our general calculations in the next section we will take all forces into account ) .
next , equation ( [ sp_mass ] ) for the mass conservation of an incompressible plasma can easily be rewritten in the local form inside the area om@xmath69@xmath60 shown in fig .
[ fig_reconnection_layer ] as @xmath70 .
thus , we have v_r/_o , ( v_y / y)_o= -(v_x / x)_o .
[ upsilon_estimate ] this equation can also be viewed as the first order taylor expansion of @xmath71 along the x - axis ( i.e. across the reconnection layer ) , @xmath72 , where we use the plasma incompressibility condition @xmath73 .
note that our first and second equations are local and are exactly the same as equations ( [ first_mhd ] ) and ( [ j_o ] ) in the sweet - parker reconnection model .
our third and fourth equations ( [ sp_local_energy ] ) and ( [ upsilon_estimate ] ) are also local and are the analogues of the two global sweet - parker equations ( [ sp_energy ] ) and ( [ sp_mass ] ) . using local equations instead of global ones is the first major difference between the sweet - parker and our theoretical models .
however , in our theoretical model we have an additional unknown parameter @xmath66 , which does not directly enter the classical sweet - parker reconnection model . as a result , we need one additional equation to be able to calculate the reconnection rate under the framework of our local model .
this additional equation comes from the condition that the right - hand - side of equation ( [ mhd_equation ] ) is constant not only across the reconnection layer , but also along the layer that is along the y - axis .
this condition is not explicitly used in the sweet - parker model , but it is used in our model , and this is the second major difference between the two models . to derive the additional equation , used in our reconnection model , we differentiate the left- and right - hand - sides of equation ( [ mhd_equation ] ) along the y - axis ( i.e. along the reconnection layer ) . the first order partial derivatives @xmath74 are identically zero because of the symmetry of the problem with respect to the x - axis .
the second order partial derivatives @xmath75 of the left- and right - hand - sides of equation ( [ mhd_equation ] ) result in 0&=&-2(v_y / y)_o(b_x / y)_o -_o(^2 j_z / y^2)_o + & & -2 + 2_oj_o / l^2 , [ sp_third_mhd ] where we use the fact that @xmath76 on the y - axis , we use our definitions @xmath66 and @xmath67 , and we estimate the second derivative of the current @xmath65 as @xmath77 .
now all five equations ( [ first_mhd ] ) , ( [ j_o ] ) , ( [ sp_local_energy])-([sp_third_mhd ] ) are local . combining them together ,
we easily obtain the sweet - parker formula for the reconnection velocity , given by equation ( [ sp_speed ] ) , which is naturally also local , see ref .
can be estimated as @xmath78 , the parameter @xmath67 can be estimated as @xmath79 , and our equation ( [ sp_third_mhd ] ) reduces to the sweet - parker equation ( [ first_mhd ] ) , as one expects . ] .
the reader of this paper could question why we develop and suggest a new theoretical approach to the problem of quasi - stationary magnetic reconnection if we obtain the same results in the sweet - parker reconnection case .
the answer is that our approach , based on local calculations , allows us to calculate the reconnection rate in the case of anomalous resistivity and also provides an additional important understanding of the reconnection problem .
our derivations for anomalous reconnection are given in the next section and we discuss our results in sec .
[ conclusions ] .
note that our local - equations approach to the reconnection problem and the more conventional global - equations approach are the same from the point of view of the underlying physics .
indeed , it is well known that by using the gauss - ostrogradski theorem , most physical equations can be written in two equivalent forms , in the form of local differential equations and in the form of global integral equations .
in this section we study reconnection with anomalous resistivity and derive a simple and accurate estimate of the reconnection rate in the classical sweet - parker - petschek two - and - a - half dimensional reconnection layer shown in fig .
[ fig_reconnection_layer ] .
we consider resistivity to be a given arbitrary function of the z - component of the electric current and the two - dimensional coordinates , @xmath80 , which has finite derivatives in @xmath1 up to the second order and in @xmath0 and @xmath65 up to the first order to be a function of @xmath65 instead of the total current @xmath81 .
this is because the reconnection process proceeds due to the z - component of the electric field , see eq .
( [ mhd_equation ] ) , and it is reasonable to assume that the electrical conductivity in the z - direction can be reduced by plasma instabilities due to large values of @xmath65 . ] .
let us list the assumptions that we make .
first , we assume that the characteristic lundquist number of the problem is large , which ( by our definition ) is equivalent to the assumption that resistivity is negligible outside the reconnection layer and the non - resistive mhd equations apply there .
second , we assume that the plasma flow is incompressible , @xmath82 . in the limit of very high lundquist numbers and slow reconnection rates
the incompressibility condition is a very good approximation in a reconnection layer even for compressible plasmas @xcite .
third , we assume that the reconnection process is quasi - stationary .
this can only be the case if the reconnection rate is small , @xmath83 [ @xmath84 and see eq .
( [ first_mhd ] ) ] , and there are no plasma instabilities in the reconnection layer .
note that in our model the reconnection rate can still be much faster than the sweet - parker rate .
fourth , we assume that the reconnection layer is thin , @xmath85 .
if plasma kinematic viscosity is small ( in comparison with resistivity ) , then the plasma outflow velocity in the downstream regions is equal to the alfven velocity @xmath15 , we have plasma mass conservation condition @xmath86 , and this assumption of a thin reconnection layer is fully equivalent to the previous assumption of a small reconnection rate . however , if plasma is very viscous , then the plasma outflow velocity is smaller than @xmath15 and assumption @xmath85 is stronger than assumption @xmath87 . finally , note that we make no assumptions about the values of the guide field @xmath88 and the plasma viscosity .
[ however , below we will see that our assumption of a quasi - stationary reconnection process in a thin current sheet layer results in a necessary condition @xmath89 for the plasma kinematic viscosity @xmath90 . refer to eq .
( [ nu ] ) for more details . ] now note that several equations that we derived in the previous section for the case of the sweet - parker reconnection with constant resistivity stay the same in the case of anomalous resistivity .
indeed , equation ( [ mhd_equation ] ) stays valid when the resistivity @xmath21 is not constant .
therefore , equation ( [ first_mhd ] ) also stays valid , except that @xmath91 now is the value of resistivity at the reconnection layer central point o ( see fig . [ fig_reconnection_layer ] ) , i.e. @xmath92 .
equations ( [ j_o ] ) and ( [ upsilon_estimate ] ) , which result from the ampere s law and the plasma incompressibility respectively , obviously stay valid too ) and ( [ upsilon_estimate ] ) are exact for the harris model reconnection sheet @xcite , which has @xmath93 , @xmath94 , @xmath95 , @xmath96 and @xmath97 . ] . at the same time , in the general case of anomalous resistivity that we consider in this section , we need to re - derive equations ( [ sp_local_energy ] ) and ( [ sp_third_mhd ] ) , which are the equations of the plasma acceleration and spatial homogeneity of the electric field z - component along the reconnection layer ( i.e. along the y - axis ) .
however , before we re - derive these two equations , we would first like to derive the equation of magnetic energy conservation . using eqs .
( [ first_mhd ] ) , ( [ j_o ] ) and ( [ upsilon_estimate ] ) , we immediately obtain b_m^2=_oj_o^2 , 1 , [ b_energy ] where for the purpose of comparison of our theoretical results to our numerical simulations we introduce a coefficient @xmath98 , which is of order of unity .
equation ( [ b_energy ] ) is the equation for magnetic energy conservation .
the rate of the supply of magnetic energy into the reconnection layer is equal to the rate of its ohmic dissipation inside the layer .
next we derive the equation of plasma acceleration along the reconnection layer ( i.e. along the y - axis ) , taking into consideration all forces acting on the plasma .
the mhd equation for the y - component of the plasma velocity @xmath64 , assuming the quasi - stationarity of the reconnection ( @xmath99 ) and plasma incompressibility ( @xmath82 ) , is @xcite ( * v*)v_y & = & -(/y ) ( p+b_x^2/2+b_y^2/2 ) + & & + ( * b*)b_y+^2v_y , [ v_y_equation ] where @xmath100 is the plasma density , @xmath90 is the plasma kinematic viscosity ( assumed to be constant ) and @xmath101 is the sum of the plasma pressure and the guide field pressure @xmath102 . taking the first order partial derivative @xmath74 of equation ( [ v_y_equation ] ) at the central point o , we obtain ^2 = -(^2 p / y^2)_o+j_o+ _ o , [ v_y_equation_2 ] where we use parameter @xmath66 and the ampere s law @xmath103 at point o , and we also use the formulas @xmath104 on the y - axis , and @xmath43 at point o , which follow from the symmetry of the problem with respect to the x- and y - axes .
the pressure term @xmath105 on the right - hand - side of equation ( [ v_y_equation_2 ] ) can be precisely calculated in analogy with the sweet - parker derivation of the pressure decline along the reconnection layer , which employs the force balance condition for the plasma across the reconnection layer and leads to equation ( [ sp_energy ] ) .
the viscosity term @xmath106_o$ ] in equation ( [ v_y_equation_2 ] ) can be calculated approximately by using estimates for the @xmath64 velocity derivatives . in appendix [ appendix_a ]
we carry out these calculations and show that the pressure and viscosity terms are equal to ( ^2 p / y^2)_o & = & b_m(^2 b_y / y^2)_m + & & + o\{^2,j_o,/_o^2 } , [ pressure_term ] + _ o & & -/_o^2 , [ viscosity_term ] where @xmath107 is calculated just outside the reconnection layer at point m ( see fig . [ fig_reconnection_layer ] ) , and expression @xmath108 denotes terms that are small compared to either @xmath109 or @xmath110 or @xmath111 in the limit of a slow reconnection rate in a thin reconnection layer , see ref .
terms can still be much larger than the @xmath107 term in equation ( [ pressure_term ] ) , in which case the pressure term @xmath105 is unimportant and negligible in equation ( [ v_y_equation_2 ] ) .
this happens when reconnection with anomalous resistivity is much faster than sweet - parker reconnection . ] .
substituting equations ( [ pressure_term ] ) and ( [ viscosity_term ] ) into equation ( [ v_y_equation_2 ] ) and using equation ( [ j_o ] ) for @xmath20 , we obtain ^2&&-b_m(^2 b_y / y^2)_m + j_o - j_o^2/b_m^2 , [ second_mhd ] + & & ( b_x / y)_o . note that this equation is exact if plasma viscosity can be neglected ( @xmath112 ) .
now we use the condition of spatial homogeneity of the electric field z - component along the reconnection layer , i.e. along the y - axis .
we take the second order partial derivatives @xmath75 of the left- and right - hand - sides of equation ( [ mhd_equation ] ) at the central point o ( note that the first order partial derivatives are identically zero ) .
taking into account the symmetry of the problem , so that @xmath104 on the y - axis , and @xmath113 at point o , we obtain & & -(^2 j_z / y^2)_o - j_o(^2/y^2)_o + & & = 2(v_y / y)_o(b_x / y)_o= 2 , [ third_mhd ] where we use formulas @xmath67 and @xmath66 .
finally , we need to estimate the @xmath114 term , which enters the left - hand - side of equation ( [ third_mhd ] ) . this estimation can be done by taking the second order partial derivative @xmath75 of equation ( [ j_o ] ) , while keeping @xmath20 constant because the partial derivative in @xmath1 is to be taken at a constant value @xmath115 . in appendix [ appendix_b ]
we give the detailed derivations and find that the y - scale of the current @xmath65 is about the same as the y - scale of the outside magnetic field , i.e. @xmath116 .
however , for the purpose of comparison of our theoretical results to numerical simulations in sec .
[ simulations ] , we find it convenient to write j_o^-1(^2 j_z / y^2)_o= b_m^-1(^2 b_y / y^2)_m , 1 , [ j_yy ] where we introduce the coefficient @xmath117 , which is of order unity .
let us take the dimensionless coefficients @xmath98 and @xmath117 , which enter equations ( [ b_energy ] ) and ( [ j_yy ] ) , and are of order unity , to be exactly unity , @xmath118 and @xmath119 .
now we have all the equations necessary to determine all unknown physical parameters .
in particular , using eqs .
( [ b_energy ] ) , ( [ second_mhd ] ) , ( [ third_mhd ] ) and ( [ j_yy ] ) , we easily obtain the following approximate algebraic equation for the z - current @xmath44 at the reconnection layer central point o : & & 3 + + + & & -(1 + ) , [ rate ] where the alfven velocity @xmath15 is defined as @xmath84 and @xmath92 is the resistivity at point o. given the resistivity function @xmath80 , as well as the magnetic field @xmath120 and its second order derivative @xmath121 outside the reconnection layer , we can solve equation ( [ rate ] ) for the current @xmath44 and find the reconnection rate , which is the rate of destruction of magnetic flux at point o , equal to @xmath122 . using eq .
( [ mhd_equation ] ) , we find that the reconnection rate is equal to @xmath123 .
note that for the classical reconnection layer that we consider ( see fig .
[ fig_reconnection_layer ] ) the right - hand - side of equation ( [ rate ] ) is positive because @xmath124 , where @xmath5 is the global scale of the magnetic field outside the reconnection layer .
once the current @xmath44 is calculated by means of equation ( [ rate ] ) , we can easily calculate all other reconnection parameters , using eqs .
( [ first_mhd ] ) , ( [ j_o ] ) , ( [ upsilon_estimate ] ) and ( [ second_mhd ] ) , v_r & & _ oj_o / b_mv_a , [ v_r ] + & & _ oj_o^2/b_m^2 , [ upsilon ] + & & j_oj_o , [ beta ] + _ o & & b_m / j_o v_r/. [ delta ] equations ( [ rate])-([delta ] ) are the most general result for magnetic reconnection that we obtain in this paper . restoring coefficients @xmath125 and @xmath126 , equation ( [ rate ] ) becomes & & ( + 2)+ + + & & = -^2(+ ) .
[ rate_general ] hereafter we will consider the natural case when @xmath127 and @xmath128 because plasma conductivity decreases as the current increases and we are interested in fast anomalous reconnection ( i.e. faster than the sweet - parker reconnection ) . in this case
the first , second and third terms on the left - hand - side of equation ( [ rate ] ) are all positive .
it is easy to see that the first term is related to sweet - parker reconnection with constant resistivity equal to @xmath91 , the second term is related to fast reconnection associated with the dependence of anomalous resistivity on the current , and the third term is related to fast reconnection associated with an _ ad hoc _ localization of resistivity in space ( see the next section for details ) .
also note that if the plasma kinematic viscosity @xmath90 is larger than the resistivity @xmath91 , then , according to equation ( [ rate ] ) , the current @xmath44 and reconnection rate @xmath129 become smaller as @xmath90 grows , i.e. the reconnection slows down for viscous plasmas as one expects .
we postpone the analysis of equations ( [ rate])-([delta ] ) until sec .
[ conclusions ] .
let us now make an estimate of the half - length of the reconnection layer @xmath4 ( see fig . [ fig_reconnection_layer ] ) .
note that @xmath4 is not needed for the calculation of the reconnection rate @xmath129 by means of equation ( [ rate ] ) . nevertheless , we are still interested in a rough estimate of @xmath4 , in particular , because we need to check our assumption that the reconnection layer is thin , i.e. that the condition @xmath130 is satisfied .
it is clear that @xmath4 can not be much larger than the global scale of the magnetic field outside the layer @xmath5 .
therefore we have the condition @xmath131 .
however , @xmath4 can be much smaller than @xmath5 , in which case the reconnection layer has a pair of the petschek switch - off mhd shocks attached to each end of the layer in the downstream regions @xcite , as shown in fig .
[ fig_reconnection_layer ] . in this case
@xmath4 should be calculated as the y - coordinate of the point on the y - axis at which the perpendicular field @xmath14 is strong enough to support the shocks @xcite .
following kulsrud @xcite , we use the jump condition on the petschek switch - off shocks to obtain ) was used by kulsrud @xcite for the case of a viscosity - free plasma .
it can be shown from the full non - ideal mhd equations that this condition is unchanged in the case of a viscous plasma . ]
v_rb_x(0,y = l)/l/= lv_a / b_m,[schock_balance ] where we use the first - order taylor expansion for an estimate @xmath132 and , as before , @xmath84 .
now , using eqs .
( [ v_r ] ) , ( [ beta ] ) and ( [ schock_balance ] ) , we can easily find @xmath4 . before we write the explicit formula for @xmath4 , note that the absolute value of the @xmath133 term in equation ( [ beta ] ) is equal or smaller than the @xmath134 term .
this follows from the fact that the left - hand - side of equation ( [ rate ] ) is equal or greater than unity [ see our comments in the paragraph that follows eq .
( [ rate_general ] ) ] .
therefore , the @xmath133 term can be omitted in equation ( [ beta ] ) for the purpose of estimating @xmath4 . as a result , we obtain the following rough estimates for the reconnection layer half - length @xmath4 and the velocity of plasma outflow in the downstream regions @xmath55 : l & & ( v_ab_m^2/_oj_o^2)(1+/_o)^-1 + & & ( v_a/)(1+/_o)^-1 , _ o ll , [ l ] + v_out & & lv_a(1+/_o)^-1 v_a , [ v_out ] where we use equations ( [ v_r ] ) , ( [ upsilon ] ) , ( [ beta ] ) and ( [ schock_balance ] ) .
note that the condition @xmath131 is always satisfied because @xmath135 , and the left- and right - hand - sides of equation ( [ rate ] ) are equal or greater than unity .
however , the condition that the reconnection layer is thin , @xmath130 , is satisfied only if plasma viscosity is not too large , _ o(v_ab_m/_oj_o)_o(v_a / v_r ) , [ nu ] where , to derive this formula , we use eqs .
( [ v_r ] ) , ( [ delta ] ) and ( [ l ] ) . in other words , to be able to form a thin reconnection layer , the plasma should not be too viscous .
note that in the case of constant resistivity and large viscosity , @xmath136 and @xmath137 , the reconnection velocity is @xmath138 ( see @xcite ) and condition ( [ nu ] ) reduces to @xmath139 .
finally , if the plasma viscosity is small in comparison to the resistivity , @xmath140 , then from equation ( [ v_out ] ) we immediately find an important and well - known result that in this case the velocity of the plasma outflowing in the downstream regions at the ends of the reconnection layer is approximately equal to the alfven velocity , @xmath141 ( see fig . [ fig_reconnection_layer ] ) . at the end of this section
we would like to discuss several assumptions that we used in our derivations .
first , the solution of equation ( [ rate ] ) is valid only if it gives @xmath142 , which is our assumption of a slow quasi - stationary reconnection .
because of equation ( [ v_r ] ) , condition @xmath142 is equivalent to @xmath17 , i.e. the reconnection velocity , which is the velocity of the incoming plasma , must be small in comparison to the alfven velocity in the upstream region .
second , the coefficient @xmath66 , given by eq .
( [ beta ] ) , must be much smaller than the current @xmath143 because the reconnection layer is assumed to be thin .
it is easy to see that this condition is satisfied .
indeed , the first term in the brackets @xmath144 $ ] in equation ( [ beta ] ) is much smaller than unity because of the upper limit for plasma viscosity given by equation ( [ nu ] ) .
the second term in the brackets in equation ( [ beta ] ) is also much smaller than unity because @xmath44 is much larger than the electric current outside the reconnection layer due to our assumption of a large characteristic lundquist number of the system .
in this section we focus on three special cases for the reconnection rate , which arise when one of the three terms on the left - hand - side of equation ( [ rate ] ) dominates over the other two .
we consider the classical sweet - parker - petschek reconnection layer shown in fig .
[ fig_reconnection_layer ] and define the global scale of the magnetic field outside the reconnection layer as @xmath145 ( see we use formula @xmath146 for the field y - component along the interval m@xmath69 that is outside the reconnection layer as shown in fig .
[ fig_reconnection_layer ] . ] ) .
in addition , for the purpose of clarity , in this section we focus only on resistivity effects and neglect plasma viscosity , assuming that @xmath140 .
first , consider the case when resistivity is constant , @xmath147 . in this case
only the first term on the left - hand - side of equation ( [ rate ] ) is nonzero and equations ( [ rate])-([delta ] ) and ( [ l ] ) reduce to lcl 3 ^ 1/2_o j_o^2l / v_ab_m^2 & & j_o(b_m / l)s_o^1/2 , + s_ov_al/_o1 , & & v_rv_as_o^-1/2 , + v_a / l , & & ( b_m / l)s_o^-1/2 , + _ ols_o^-1/2 , & & ll , [ sp_reconnection ] where we set the plasma kinematic viscosity to zero ( @xmath112 ) , use our definition of the global field scale @xmath145 , introduce the lundquist number @xmath148 and assume for our estimates that @xmath149 . the above equations are the sweet - parker reconnection equations with constant resistivity equal to @xmath91 .
thus , we find that if resistivity is constant , then the reconnection must be sweet - parker and not petschek @xcite .
we discuss this important result in sec .
[ conclusions ] .
note that in this section , contrary to sec .
[ sweet_parker_model ] , we do not assume the sweet - parker geometrical configuration for the reconnection layer , but derive it together with the reconnection rate from our general equations of the previous section .
a typical configuration of the reconnection layer in the case of sweet - parker reconnection is shown on the left - bottom plot in fig .
[ fig_rosette ] .
this plot is marked by letters `` s - p '' .
now let us consider the case when resistivity is anomalous and is a monotonically increasing function of the electric current only , @xmath150 .
let us further assume that this dependence of resistivity on the current is very strong , so that @xmath151 . in this case the second term on the left - hand - side of equation ( [ rate ] ) is dominant , and equations ( [ rate ] ) and ( [ l ] ) reduce to ( d / d j_z)_o & & _ o^3j_o^3l^2/v_a^2b_m^4 + & & ^1/3 , [ pk_reconnection_rate ] + l & & l + & & l^-1/2 l. [ pk_reconnection_l ] here again we set @xmath112 , and we use the formula @xmath152 and eqs .
( [ v_r ] ) and ( [ delta ] ) . from equation ( [ pk_reconnection_l ] ) we see that the half - length of the reconnection layer @xmath4 is much less than the global field scale @xmath5 in the case of a strong dependence of resistivity on the current ( @xmath151 ) .
this means that in this case the geometry of the reconnection layer is petschek with a pair of shocks attached to each end of the layer in the downstream regions , as shown in fig .
[ fig_reconnection_layer ] .
equation ( [ pk_reconnection_rate ] ) for the reconnection rate was first analytically derived by kulsrud @xcite for a special case when @xmath153 . in the petschek - kulsrud reconnection case
the reconnection rate , given by eq .
( [ pk_reconnection_rate ] ) , can be considerably faster than the sweet - parker reconnection rate . this fast reconnection has been previously observed in many numerical simulations done with anomalous resistivity @xmath150 ( e.g. see @xcite ) .
the typical configuration of the reconnection layer in the case of petschek - kulsrud reconnection is shown in the left - bottom plot in fig .
[ fig_rosette ] , this plot is marked by letters `` p - k '' .
finally let us consider the special case when resistivity is given by @xmath154 and it is spatially localized around the central point o of the reconnection layer ( see fig . [ fig_reconnection_layer ] ) , so that @xmath155 .
in other words , we assume that resistivity is anomalous and is localized on scale @xmath156 that is much smaller than the global field scale @xmath5 . in this case the third term on the left - hand - side of equation ( [ rate ] ) is dominant , and equations ( [ rate])-([delta ] ) and ( [ l ] ) reduce to lcl l / l__o j_o^2l / v_ab_m^2 & & j_o(b_m / l_)s_l^1/2 , + s_lv_al_/_o1 , & & v_rv_as_l^-1/2 , + v_a / l _ , & & ( b_m / l_)s_l^-1/2 , + _ ol_s_l^-1/2 , & & ll_l , [ localized_reconnection ] where we again set @xmath112 .
the above equations are the same as the sweet - parker equations ( [ sp_reconnection ] ) with the global field scale @xmath5 replaced by the resistivity scale @xmath157 .
note that , when resistivity is localized , the reconnection rate becomes faster than the sweet - parker rate by a factor @xmath158 and the geometry of the reconnection layer is petschek with a pair of shocks attached to each end of the layer ( see fig . [ fig_reconnection_layer ] ) .
these results are in agreement with many previous numerical simulations of reconnection with spatially localized resistivity @xcite .
in this section we present the results of our numerical simulations of unforced reconnection of two cylindrical magnetic flux tubes .
these simulations are not intended as a check or a proof of our theoretical results for magnetic reconnection .
our equations ( [ rate])-([delta ] ) have been derived analytically and are very general .
a comprehensive testing of them would require extensive computational work , which is beyond the scope of this paper .
instead , we present our simulations as a demonstration of our reconnection model predictions . following kulsrud @xcite , we assume that plasma resistivity is given by the following piecewise linear function of the z - component of the electric current : ( j_z)=_s+_\{0,j_z - j_c}/j_c , [ linear_resistivity ] where @xmath159 is the spitzer resistivity , which is assumed to be very small , @xmath160 is the anomalous resistivity parameter and @xmath161 is the critical current parameter .
the kulsrud model s prediction for the reconnection rate in the case @xmath162 , which is given by equation ( [ pk_reconnection_rate ] ) , has already been checked and confirmed numerically by breslau and jardin @xcite . here
we simulate reconnection with anomalous resistivity given by equation ( [ linear_resistivity ] ) for a different computational setup , a higher lundquist number and a wider range of parameters @xmath160 and @xmath161 ( without the restriction @xmath163 ) .
our intent is to see how our general formula ( [ rate ] ) for the reconnection rate works in this case .
we consider an unforced reconnection of two cylindrical magnetic flux tubes with the initial z - component of the magnetic field vector potential equal to a_z(x , y)&=&a_0 , + a_0&=&b_0r_0 , r_^2=(xd)^2+y^2 , [ tubes ] see the top - left plot in fig .
[ fig_rosette ] .
this convenient computational setup was suggested to us by mikic and vainshtein @xcite .
we choose the parameters in equation ( [ tubes ] ) as @xmath164 ( the global scale of the field is unity ) , @xmath165 ( @xmath166% of initially reconnected flux ) and @xmath167 ( the maximal initial field is unity ) .
thus , @xmath168 .
in addition , for convenience we choose the plasma density @xmath169 , so that the typical alfven velocity and time are unity , @xmath170 and @xmath171 . the guide field is chosen to be zero , @xmath18 , and the initial plasma velocities are zero .
the initial gas pressure @xmath101 is chosen in such way that each of the two cylindrical magnetic flux tubes ( [ tubes ] ) would initially be in complete equilibrium , @xmath172 , if there were no magnetic forces from the other tube .
the plasma kinematic viscosity is chosen to be equal to the spitzer resistivity , @xmath173 .
the boundary conditions are placed at @xmath174 , which are virtually at infinity ( the magnetic vector potential ( [ tubes ] ) drops to less than @xmath175 at the boundaries ) . because of the symmetry of the problem , in the case of a quasi - stationary reconnection considered here , it is enough to run simulations only in the upper - right quadrangle of the full computational box .
we use the flash code for our simulations .
this is a compressible adaptive - mesh - refinement ( amr ) code written and supported at the asc center of the university of chicago .
( for a comprehensive description of the flash code see @xcite ) .
the mhd module of the code uses central finite differences to properly resolve all resistive and viscous scales . comparing numerical results obtained by simulations done with a compressible code to our approximate theoretical formulas derived for incompressible fluids
is fine in a case of a very high lundquist number .
this is because in this case the incompressibility condition is a very good approximation in a reconnection layer even for compressible plasmas @xcite .
indeed , in our simulations the plasma density varies by no more than 15% inside the reconnection layer .
the biggest advantage of the flash code for our purposes is that it is an already existing , well tested code with the amr feature , which allows us to place the boundary conditions at infinity .
the size of the smallest elementary grid cell in our two - dimensional simulations was chosen typically to vary from @xmath176 to @xmath177 , which was sufficient to resolve the resistive reconnection layer .
the two cylindrical magnetic flux tubes , initially set up according to equation ( [ tubes ] ) , attract each other ( in a similar way as two wires with colliniar currents do ) . as a result , as time goes on , the tubes move toward each other , form a thin reconnection layer along the y - axis and eventually completely merge together by reconnection .
this merging process is displayed in the three top plots in fig .
[ fig_rosette ] , which show the field vector potential @xmath178 in a central region of the full computational box .
the two bottom - left plots in fig .
[ fig_rosette ] show the electric current @xmath65 in a central region that includes the reconnection layer .
these plots clearly show the reconnection layer configuration which is formed during the reconnection process in the cases of the sweet - parker and petschek - kulsrud reconnection ( refer to secs .
[ sweet_parker_reconnection ] and [ petschek_kulsrud_reconnection ] ) . the bottom - right plot in fig .
[ fig_rosette ] demonstrates the functional dependence on time typical of the normalized reconnection rate @xmath179 at the reconnection layer central point o. ( this point is shown in fig .
[ fig_reconnection_layer ] ) .
next we compare the maximal ( peak ) reconnection rate observed in the numerical simulations with the theoretical rate predicted by our reconnection model .
when resistivity is given by equation ( [ linear_resistivity ] ) and @xmath173 , our theoretical formula ( [ rate ] ) for the reconnection rate reduces to 3 + ( 1+_s/_o ) , [ linear_model_rate ] where , as explained above , in our numerical simulations we choose @xmath167 , @xmath164 , @xmath180 and @xmath169 , and by definition the global field scale is @xmath145 .
we also assume that @xmath181 , which is the case that we consider in our numerical simulations .
as we can see from equation ( [ linear_model_rate ] ) , the theoretical reconnection rate @xmath129 is rather sensitive to @xmath120 and @xmath5 , which are the strength and scale of the magnetic field at point m outside the reconnection layer ( see fig . [ fig_reconnection_layer ] ) .
therefore , in order to compare the theoretical results with the results of our simulations , we need to accurately calculate the @xmath120 and @xmath5 observed in the simulations . as a result ,
the choice of the exact position of point m , at which @xmath120 and @xmath5 are calculated , is important .
first , in our simulations we choose the point m to be the point on the x - axis at which the observed resistivity term @xmath182 is three times smaller than that observed at the central point o , i.e. @xmath183 .
the three plots on the left in fig .
[ fig_rates_1_3 ] demonstrate our results for this choice .
the top plot on the left shows the reconnection rate .
the crosses are a log - log plot of the maximal ( peak ) reconnection rate observed in our simulations as a function of parameter @xmath160 for fixed @xmath184 and @xmath185 .
the boxes are the theoretical reconnection rate , which is given by equation ( [ linear_model_rate ] ) with the appropriate values of @xmath120 , @xmath5 and density @xmath100 observed in the simulations .
the solid horizontal line ( simulations ) and the dashed horizontal line ( theory ) correspond to the @xmath186 case .
the inclined dotted line demonstrates the @xmath187 scaling [ refer to eq .
( [ pk_reconnection_rate ] ) ] .
the crosses and the boxes do not follow the @xmath187 scaling for large values of @xmath160 simply because in this case the reconnection rate becomes relatively fast and the magnetic field @xmath120 outside of the reconnection layer is not piled up as much as in the case when @xmath160 is small and the reconnection rate is relatively slow . as a result ,
our rate curves flatten at large values of @xmath160 .
the central and the bottom plots on the left in fig .
[ fig_rates_1_3 ] show the observed values of coefficients @xmath98 and @xmath117 , which are directly calculated from the simulation data by using equations ( [ b_energy ] ) and ( [ j_yy ] ) .
as we can see , @xmath98 and @xmath117 are of order of unity , as one expects .
the three plots on the right in fig .
[ fig_rates_1_3 ] are the same as the three plots on the left except point m is now chosen as the point on the x - axis at which @xmath188 . comparing the plots on the left and the plots on the right , we see that the choice of point m is indeed important .
note that for the choice @xmath183 for the position of point m the half - thickness of the reconnection layer @xmath20 , defined as the abscissa of point m , increases from @xmath189 to @xmath190 as @xmath160 increases from zero to its maximal value shown on the plots in fig .
[ fig_rates_1_3 ] . at the same time , for the choice @xmath188 for the position of point m @xmath20 ranges from @xmath191 to @xmath192 , which is noticeably larger .
perhaps simulations of forced reconnection with a strict control of position of point m together with control of the outside field @xmath120 and its scale @xmath5 , could be better suited for comparison to our theoretical model .
such simulations are beyond the scope of this paper .
however , see more discussion of forced reconnection in the next section .
finally , figure [ fig_rates_2_3 ] shows the same results as fig .
[ fig_rates_1_3 ] , except the former has plots for a smaller value of the critical current , @xmath193 .
we believe that the results presented in figs . [ fig_rates_1_3 ] and [ fig_rates_2_3 ] generally confirm our theoretical model .
in particular , in all cases the theoretical reconnection rates and the rates observed in the simulations do not differ by more than 33% .
the observed relatively small discrepancy between the theoretical and simulated rates is mainly due to the coefficient @xmath98 not being precisely constant in the simulations , while the variations of coefficient @xmath117 are somewhat less important [ see eq .
( [ rate_general ] ) for the theoretical reconnection rate ] .
this discrepancy can be due to two causes .
first , our theoretical model is general , but approximate , and second , the plasma is compressible in the simulations , while it is assumed to be incompressible in the theoretical model .
let us summarize our main results . in this paper we take a new theoretical approach to the calculation of the rate of quasi - stationary magnetic reconnection .
our approach is based on analytical derivations of the reconnection rate from the resistive mhd equations in a small region of space that is localized about the center of a thin reconnection layer and has its size equal to the layer thickness .
our local - equations approach turns out to be feasible and insightful .
it allows us to consider magnetic reconnection with an arbitrary anomalous resistivity and to calculate the reconnection rate for this general case [ see eq .
( [ rate ] ) ] .
we find the interesting and important result that if plasma is incompressible and reconnection is quasi - stationary , then the reconnection rate is determined by the anomalous resistivity function @xmath194 and by the strength and structure of the magnetic field just outside of the reconnection layer ( i.e. at point m in fig . [ fig_reconnection_layer ] ) .
thus , we find that the global magnetic field and its configuration are not directly relevant for the purpose of calculation of a quasi - stationary reconnection rate , although , of course , the local magnetic field outside the reconnection layer depends on the global field .
one of the major results of this paper is that in the case of constant resistivity , @xmath30 , the magnetic reconnection rate is the slow sweet - parker reconnection rate and not the fast petschek reconnection rate [ refer to eqs .
( [ sp_reconnection ] ) ] .
this result agrees with numerical simulations and at the same time contradicts the result of the original petschek theoretical model . in the framework of our theoretical approach , based on local equations ,
the reason for this contradiction can be understood as follows : in the petschek model our parameter @xmath195 , which is equal to the first order partial derivatives of the incompressible plasma velocities at the reconnection layer center , is basically treated as a free parameter .
this is because in the petschek model @xmath196 can be estimated as the ratio of the plasma outflow velocity ( equal to the alfven velocity for viscosity - free plasma ) and the reconnection layer length , @xmath197 , and the layer length @xmath4 is treated as a free parameter by petschek . in his model
@xmath4 is taken to be equal to the minimal possible value , that the petschek shocks do not seriously perturb the magnetic field in the upstream region .
this is @xmath198 , where @xmath199 is the lundquist number @xcite . in this case
@xmath200 and , according to equations ( [ v_r ] ) and ( [ upsilon ] ) , the reconnection velocity in this case is equal to @xmath201 , which is the petschek result . on the other hand , in our theoretical model
the parameter @xmath196 is not treated as a free parameter .
in fact , our three physical parameters @xmath67 , @xmath66 and @xmath202 are connected to each other and must be calculated from equations ( [ b_energy ] ) , ( [ second_mhd ] ) and ( [ third_mhd ] ) .
let us discuss the meaning of these equations .
equation ( [ b_energy ] ) is the equation of magnetic energy conservation .
it says that the rate of supply of the magnetic field energy into the reconnection layer @xmath203 must be equal to the rate of the resistive dissipation of this energy @xmath204 . in the petschek model @xmath196 and , accordingly , the rate of the magnetic energy supply @xmath203 are basically prescribed by hand ( resulting in an _ ad hoc _ fast reconnection ) , while in our model they are self - consistently calculated from the mhd equations .
equation ( [ second_mhd ] ) is the equation of plasma acceleration along the reconnection layer .
it says that the magnetic tension force @xmath110 must be large enough in order to be able to push out all the plasma along the layer that is supplied into the layer .
finally , equation ( [ third_mhd ] ) is the equation of spatial homogeneity of the electric field z - component along the reconnection layer .
this equation sets an upper limit on the product @xmath205 in the case of a quasi - stationary reconnection and is directly related to the calculations and arguments given by kulsrud @xcite in the framework of the global - equations theoretical approach , see ref .
are directly related to the explanation of why the petschek reconnection model does not work , as shown by kulsrud @xcite in the framework of the global - equations .
kulsrud s argument is that the length of petschek reconnection layer @xmath4 is not a free parameter , but must be determined by the condition that the perpendicular magnetic field component @xmath14 has to be regenerated by the rotation of the parallel field component @xmath54 at the same rate as it is being swept away by the downstream flow .
it is easy to see that the integration of `` local '' equation ( [ mhd_equation ] ) without the resistivity term over the area of the contour o@xmath206m@xmath206@xmath69@xmath206@xmath60 shown in fig .
[ fig_reconnection_layer ] will result in the same `` global '' equation for the balance of the @xmath14 field , which was used by kulsrud in his work @xcite . ] . as a result
, none of parameters @xmath196 , @xmath207 and @xmath44 can be treated as free parameters , and all of them must be self - consistently determined from the mhd equations .
it is very instructive to briefly examine our results from the two distinct points of view in connection with the reconnection problem , which are expressed in numerous papers on computer simulation of magnetic reconnection .
these two points of view are : unforced ( free ) magnetic reconnection and forced magnetic reconnection . in the case of unforced reconnection
, one should solve our main equation ( [ rate ] ) for the current @xmath44 at the reconnection layer center and then solve for the reconnection velocity @xmath16 by using equation ( [ v_r ] ) .
the solution for @xmath44 and @xmath16 will depend on @xmath120 , which is the strength of the magnetic field outside the reconnection layer and which enters the right - hand - side of equation ( [ rate ] ) . on the other hand , in the case of forced magnetic reconnection the reconnection velocity @xmath16 is prescribed and fixed . in this case
the magnetic field outside the reconnection layer @xmath120 should be treated as an unknown quantity , and equations ( [ rate ] ) and ( [ v_r ] ) should be solved together in order to find the correct quasi - stationary values of @xmath44 and @xmath120 . in other words , in the forced reconnection case an initially weak outside magnetic field @xmath120 gets piled up to higher and higher values until the resulting current @xmath44 in the reconnection layer becomes large enough to be able to exactly match the prescribed reconnection velocity @xmath16 and to be able to reconnect all magnetic flux and magnetic energy , which are supplied into the reconnection region in the quasi - stationary reconnection regime .
finally , a couple of words about plasma viscosity and guide field and their effect on magnetic reconnection .
first , according to our equations ( [ rate ] ) and ( [ v_r ] ) , in the case when the resistivity is constant , @xmath30 , and the plasma viscosity is much larger than resistivity , @xmath137 , the reconnection velocity becomes @xmath208 , which is @xmath209{\nu/\eta_o}$ ] times smaller than the sweet - parker reconnection velocity given by formula ( [ sp_reconnection ] ) , see @xcite .
thus , we see that the reconnection rate becomes smaller when the plasma viscosity becomes large . however , note that in many astrophysical and laboratory applications plasmas are very hot and highly rarefied . under these conditions
the ion gyro - radius becomes much shorter than the ion mean - free - path , and the plasma becomes strongly magnetized . as a result ,
the plasma viscosity becomes the braginskii viscosity , which is dominated by magnetized ions @xcite . in this case in all our equations above the isotropic viscosity @xmath90 , which is proportional to the ion mean - free - path , should be replaced by the braginskii perpendicular viscosity , which is proportional to the ion gyro - radius and is much smaller than the perpendicular viscosity in a strongly magnetized plasma .
second , according to our results , in two - and - a - half dimensional geometry the guide field @xmath88 has no effect on the quasi - stationary reconnection rate .
indeed , in our derivations the guide field appears only as magnetic pressure @xmath102 term in addition to the plasma pressure .
the combined pressure @xmath101 enters equations ( [ v_y_equation ] ) and ( [ v_y_equation_2 ] ) of plasma acceleration along the reconnection layer and the value of spatial derivative of @xmath101 is given by equation ( [ pressure_term ] ) , which does not involve @xmath88 .
thus , the guide field gets eliminated and does not enter into our final equation ( [ rate ] ) for the reconnection rate . however ,
if one assumes that the anomalous resistivity @xmath21 depends on x- and y - components of the current @xmath210 and @xmath211 in addition to its dependence on the z - component of the current @xmath65 , then the reconnection rate will depend on the guide field @xmath88 .
in this paper we consider quasi - stationary magnetic reconnection in a thin reconnection layer .
we leave a study of tearing modes instability in a reconnection layer and non - quasi - stationary reconnection for a future paper .
it is our special pleasure to thank ellen zweibel and dmitri uzdensky for many stimulating discussions and useful comments .
we are grateful to zoran mikic and samuel vainshtein for suggesting to us the convenient computational setup used in our simulations .
we would also like to thank andrey beresnyak , amitava bhattacharjee , stas boldyrev , fausto cattaneo , jeremy goodman , hantao ji , alexander obabko , robert rosner and masaaki yamada for a number of valuable comments .
this work was supported by the center for magnetic self - organization ( cmso ) grant .
the numerical simulations were supported by a center for magnetic reconnection studies ( cmrs ) grant .
the software used in this work was in part developed by the doe supported asc / alliances center for astrophysical thermonuclear flashes at the university of chicago .
the simulations were carried out on doe computers at the oak ridge national laboratory .
below , for brevity , we assume that spatial derivatives are to be taken with respect to all indexes that are listed after the comma signs in the subscripts , e.g. @xmath212 .
we derive equation ( [ pressure_term ] ) first .
the derivation is somewhat analogous to the sweet - parker derivation of the pressure decrease along the reconnection layer , which leads to equation ( [ sp_energy ] ) .
namely , to find the pressure decrease and the pressure second derivative along the layer , we integrate the pressure gradient vector along the contour o@xmath206m@xmath206@xmath69@xmath206@xmath60 shown in fig .
[ fig_reconnection_layer ] and use the force balance condition for the plasma across the reconnection layer .
now we carry out these calculations in a mathematically precise way . for infinitesimally small values of the y - coordinate , taking into account the symmetry of the reconnection layer with respect to the x- and y - axes and plasma incompressibility , we use the following taylor expansions in @xmath1 for the plasma velocity @xmath213 and for the magnetic field @xmath214 : l v_x = v_x^(0)(x)+(y^2/2)v_x , yy^(0)(x)+(y^4/24)v_x , yyyy^(0)(x ) , + b_x = yb_x , y^(0)(x)+(y^3/6)b_x , yyy^(0)(x ) , + v_y = -yv_x , x^(0)(x)+(y^3/6)v_y , yyy^(0)(x ) , + b_y = b_y^(0)(x)+(y^2/2)b_y , yy^(0)(x ) , [ app_a_expansions ] where the variables with the superscripts @xmath215 are taken at @xmath216 and depend only on @xmath0 . assuming quasi - stationarity of reconnection ( @xmath99 ) and plasma incompressibility , the mhd equation for the plasma velocity @xmath8 is p+**(b_x^2+b_y^2)/2&= & -(*v***)*v*+ ( * b***)*b * + & & + ^2*v * , [ app_a_v_equation ] where @xmath101 is the sum of the plasma pressure and the z - component field magnetic pressure @xmath102 .
let us calculate line integrals of the left- and right - hand - sides of equation ( [ app_a_v_equation ] ) along the contour o@xmath206m@xmath206@xmath69@xmath206@xmath60 shown in fig .
[ fig_reconnection_layer ] .
first , the line integral of the left - hand - side is obviously & & p(0,y)-p(0,0)+(1/2)b_x^2(0,y ) + & & = ( y^2/2)[p_,yy(0,0)+^2 ] , [ app_a_lhs_integral ] where @xmath58 is the y - coordinate of points @xmath69 and @xmath60 , and we use the formulas @xmath217 on the y - axis , @xmath94 at point o and @xmath61 for small @xmath1 [ see the definition of @xmath207 in eq .
( [ second_mhd ] ) ] .
second , using expansion formulas ( [ app_a_expansions ] ) , we calculate the following line integrals along the contour o@xmath206m@xmath206@xmath69@xmath206@xmath60 , up to the second order in @xmath58 : & = & ( ^2- v_x^(m ) + & & + 2_o^m v_x , x^(0)v_x , yy^(0)dx ) , [ app_a_v_integral ] + & = & ( b_y^(m)b_y , yy^(m)+^2 + b_x , y^(m)j_z^(m ) + & & -_o^m dx ) , [ app_a_b_integral ] + & = & ( 2v_y , yyy^(m)-v_x , xxx^(m)-v_y , yyy^(0)(0 ) + & & -_o^m v_x , yyyy^(0)dx ) , [ app_a_nu_integral ] where the variables with the superscripts @xmath218 are calculated at point m ( see fig . [ fig_reconnection_layer ] ) .
next we estimate the terms on the right - hand - sides of eqs .
( [ app_a_v_integral])-([app_a_nu_integral ] ) .
recall the notations @xmath4 and @xmath20 for the half - length and half - thickness of the reconnection layer ( see fig .
[ fig_reconnection_layer ] ) , and that @xmath85 ( which is our assumption of a thin reconnection layer ) .
the z - current inside the reconnection layer is approximately equal to @xmath219 [ see eq .
( [ j_o ] ) ] , where @xmath220 is the field at point m , while the z - current outside the layer is @xmath221 .
the last two terms on the right - hand - side of equation ( [ app_a_b_integral ] ) can be estimated as @xmath222 and @xmath223dx \sim\delta_o b_m\beta / l'^2\sim\beta j_o(\delta_o / l')^2\ll \beta j_o$ ] .
next , from equation ( [ mhd_equation ] ) with the resistivity term dropped we see that outside the reconnection layer the typical scale of the plasma velocity @xmath8 is about the same as the typical scale of the magnetic field @xmath9 and , therefore , can not be smaller than @xmath4 .
thus , the estimates of the three terms on the right - hand - side of equation ( [ app_a_v_integral ] ) are @xmath224 ^ 2\sim v_x^{(m)}[v_{x , xx}^{(m)}+v_{x , yy}^{(m)}]\sim v_r^2/l'^2\sim\upsilon^2(\delta_o / l')^2\ll \upsilon^2 $ ] [ see eq .
( [ upsilon_estimate ] ) ] and @xmath225 .
the estimates of the four terms on the right - hand - side of equation ( [ app_a_nu_integral ] ) are @xmath226 , @xmath227 and @xmath228 .
note that here we use @xmath4 for estimation of y - derivatives .
in fact , using the global scale @xmath5 of the magnetic field outside the reconnection layer would have been more appropriate for some of the estimations ( as shown in appendix [ appendix_b ] ) .
however , using @xmath229 for the upper estimates of the @xmath74 derivatives is perfectly fine for the purposes in this appendix .
next , calculating the line integral of the right - hand - side of eq .
( [ app_a_v_equation ] ) by using formulas ( [ app_a_v_integral])-([app_a_nu_integral ] ) and our estimates made in the previous paragraph , and then equating the result to the right - hand - side of eq .
( [ app_a_lhs_integral ] ) , we easily obtain equation ( [ pressure_term ] ) .
note that the first term on the right - hand side of equation ( [ pressure_term ] ) can also be written in terms of the second derivative of the magnetic pressure at point m : @xmath230 ( note that @xmath231 because the reconnection layer is thin ) .
therefore , as noted by zweibel @xcite , equation ( [ second_mhd ] ) is similar to bernoulli s equation .
now we derive equation ( [ viscosity_term ] ) , which gives an approximate estimate of the viscosity term @xmath106_o$ ] in equation ( [ v_y_equation_2 ] ) .
we make this estimate as follows : note that @xmath232 on the y - axis because of the symmetry of the problem relative to this axis .
therefore , from the second order taylor expansion of @xmath233 in @xmath0 , we obtain an approximate formula @xmath234/\delta_o^2 \approx - v_{y , y}(0,0)/\delta_o^2=-\upsilon/\delta_o^2 $ ] , where we take into account that @xmath235 .
we can rewrite this approximate formula for @xmath236 as the following exact formula : @xmath237 , where @xmath238 is an unknown coefficient of order unity .
our numerical simulations of reconnection with constant resistivity show that @xmath238 is indeed about unity if @xmath20 is estimated by equation ( [ j_o ] ) .
thus we immediately find that @xmath106_o= \rho\nu[v_{y , yxx}(0,0)+v_{y , yyy}(0,0)]\approx \rho\nu v_{y , yxx}(0,0)\approx-\rho\nu\upsilon/\delta_o^2 $ ] , which is equation ( [ viscosity_term ] ) .
here we also use @xmath239 .
here as in appendix [ appendix_a ] , we assume that spatial derivatives are to be taken with respect to all indexes that are listed after the comma signs in the subscripts , e.g. @xmath240 .
we derive equation ( [ j_yy ] ) in two steps .
first , we estimate @xmath241 at points o and m ( see fig . [ fig_reconnection_layer ] ) , since we will need these estimates below .
consider the formula @xmath242 , which represents the fact that the magnetic field is divergence - free .
take the @xmath74 and @xmath243 derivatives of this formula and integrate the resulting equations along the interval om shown in fig .
[ fig_reconnection_layer ] .
we obtain b_x , y^(m)-&= & _
b_x , yxdx = -_o^m b_y , yydx + & & -_o b_y , yy^(m ) , [ app_b_bxy ] + b_x , yyy^(m)-b_x , yyy^(o ) & = & _ o^m b_x , yyyxdx = -_o^m b_y , yyyydx + & & -_o b_y , yyyy^(m ) , [ app_b_bxyyy ] where the variables with the superscripts @xmath244 and @xmath218 are taken at points o and m respectively and @xmath245 [ see eq .
( [ second_mhd ] ) ] .
note that @xmath20 is the half - thickness of the reconnection layer , equal to the abscissa of point m ( see fig .
[ fig_reconnection_layer ] ) . in making the estimates of the integrals on the right - hand - sides of eqs .
( [ app_b_bxy ] ) and ( [ app_b_bxyyy ] ) , we take into account that @xmath246 because @xmath217 on the y - axis . now , using eq .
( [ app_b_bxy ] ) , we estimate that @xmath247 .
let @xmath5 be the global scale of the magnetic field outside the reconnection layer . then , since point m is located outside the reconnection layer , we have @xmath248 and @xmath249 .
next , using these estimates , the above estimate for @xmath250 and eq .
( [ app_b_bxyyy ] ) , we obtain the following formula : b_x , yyy^(o)~b_x , yyy^(m)~/l^2(_o / l^2)b_y , yy^(m ) . [ app_b_bx_estimates ]
second , we estimate @xmath251 at point o. consider ampere s law formula @xmath252 .
we take the @xmath75 derivative of this equation and integrate the result along the interval om shown in fig .
[ fig_reconnection_layer ] .
we find that _
o^mj_z , yydx & = & _ o^mb_y , yyxdx-_o^mb_x , yyydx + & = & b_y , yy^(m)-_o^mb_x , yyydx
. the integral @xmath253 can be estimated by using formula ( [ app_b_bx_estimates ] ) .
the integral of @xmath251 can be estimated as @xmath254 . as a result
, we obtain _ o j_z , yy^(o ) b_y , yy^(m)_o _
ob_y , yy^(m)b_y , yy^(m ) , [ app_b_app_b_jzyy ] where we use @xmath31 and latexmath:[$\delta_o\beta / l^2=(\delta_o j_o / l^2)(\beta / j_o)\approx ( b_m / l^2)(\beta / j_o)\approx \pm b_{y , yy}^{(m)}(\beta / j_o)\ll as suggested by uzdensky @xcite , there exists a nice graphical interpretation of the fact that the y - scale of the current @xmath65 is about the same as the scale of the outside magnetic field , i.e. @xmath258 and that equation ( [ j_yy ] ) holds .
there can be two different cases of the reconnection layer geometry .
first , the half - length of the reconnection layer @xmath4 can be approximately equal to the global scale @xmath5 of the outside field . in this case
the reconnection is sweet - parker and @xmath19 is the only available scale in the y - direction .
therefore , in this case @xmath259 . in the second case
the reconnection layer half - length is much smaller than the global scale , @xmath260 , and the reconnection is fast ( relative to the sweet - parker reconnection ) . in this case , the z - current @xmath261 on the y - axis drops abruptly , as the y - coordinate passes value @xmath4 and point @xmath262 moves from the region inside the reconnection layer to the region of the outflowing plasma that is located between the petschek shocks ( see fig .
[ fig_reconnection_layer ] ) . however , the z - current @xmath65 stays large inside the shocks .
in other words , @xmath65 is a smooth function ( on the global scale @xmath5 ) along the lines that lie inside the reconnection layer and extend into the shock separatrices .
thus , in this case , despite @xmath260 , the y - scale of the z - current at the reconnection layer central point o is still @xmath5 .
this graphical interpretation is well demonstrated by the bottom - left plot of the current for the petschek - kulsrud ( `` p - k '' ) reconnection case in fig .
[ fig_rosette ] . | in this paper quasi - stationary , two - and - a - half - dimensional magnetic reconnection is studied in the framework of incompressible resistive magnetohydrodynamics ( mhd ) . a new theoretical approach for calculation of the reconnection rate
is presented .
this approach is based on local analytical derivations in a thin reconnection layer , and it is applicable to the case when resistivity is anomalous and is an arbitrary function of the electric current and the spatial coordinates .
it is found that a quasi - stationary reconnection rate is fully determined by a particular functional form of the anomalous resistivity and by the local configuration of the magnetic field just outside the reconnection layer .
it is also found that in the special case of constant resistivity reconnection is sweet - parker and not petschek . |
understanding the origin of the initial stellar mass function ( imf ) remains one of the most challenging issues in modern astrophysics .
when averaged over the total volume of galaxies or whole stellar clusters , the imf is observed to follow a nearly uniform behavior which consists in an increased number of stars counted when going from the most massive stars up to @xmath2 @xmath1 , followed by a shallower increase between @xmath2 and @xmath3 @xmath1 and a decline in the number of stars at masses @xmath4 @xmath5 .
this standard imf has been described , with continuous refinements , by several analytical functions ( e.g. , salpeter 1955 ; miller - scalo 1979 ; kroupa 2002 ; chabrier 2003 ) . yet
, deviations from the standard imf at low and high mass ends have been reported in many observations ( see review in elmegreen 2004 ) . at high mass ,
the imf is observed to be generally top - heavy in dense cluster cores such as in the arches cluster ( e.g. , stolte et al . 2005 ; kim et al .
2006 ) and stars appear to be , preferentially located in the central parts of the clusters ( e.g. , hillenbrand & hartmann 1998 ; figer et al .
1999 ; stolte 2002 ; gouliermis et al .
star - bursts regions are also observed to possess a top - heavy imf , either in the form of a shallow slope at high mass ( e.g. , einsenhauer et al .
1998 ; sternberg 1998 ) or by having a value of the high mass - low mass turnover of a few to several @xmath5 which is substantially larger than that of the standard imf ( e.g. , rieke 1993 ) .
the imf of dense clusters seems also to be truncated at the very high mass end ( e.g. , stolte 2005 ) .
the mass truncation can be attributed to the short lifetimes of the most massive stars .
ideas that have been proposed to explain the shallowness of the slope at the high mass end include a ) a model based on the coalescence of pre - stellar cores ( pscs ) and their subsequent gravitational collapse to produce stars ( e.g. , nakano 1966 ; silk & takahashi 1979 ; elmegreen & shadmehri 2003 ; elmegreen 2004 ; shadmehri 2004 ) , b ) the mass segregation of stars in the cluster ( e.g. , vesperini & heggie 1997 ; kroupa 2002 ; mouri & taniguchi 2002 ) , and c ) a renewed episode of gas accretion by the cluster under favorable conditions , which leads to the formation of a new generation of massive stars ( e.g. , lin & murray 2007 ) .
this latter idea is somehow inconsistent with the fact that a cluster such as the arches cluster is overall very young ( i.e , age @xmath6 myrs ) and may apply only to older clusters . concerning mass segregation , whereas there is little doubt that the enhancement in the numbers of massive stars in the inner parts of the cluster by dynamical processes will lead to a shallower imf , this does not constitute a direct proof that the primordial imf of stars in those regions was not shallower than a salpeter imf initially .
the latter is commonly used as an initial input for the stellar distribution functions at all cluster radii in n - body models ( e.g. , portegies zwart et al .
furthermore , the imf of the arches cluster is characterized by a peculiar bump at @xmath7 @xmath1 which is not , to date , well reproduced by the effect of mass segregation in n - body simulations ( e.g. , kim et al .
2006 , portegies zwart et al .
2007 ) . in this letter
, we propose a coalescence model in which the local initial pscs populations are those resulting from the local gravo - turbulent fragmentation of the protocluster cloud .
we follow the time evolution of the mass function of pscs and the transition to the imf under the assumption that the coalescence of pscs is important .
this is very likely to be the case for the pscs located in the central parts of the protocluster cloud .
we consider pscs ( e.g. , andr et al . 2000 ) embedded in an isothermal mc ( at a temperature of @xmath8 k ) , at different locations @xmath9 from the cloud s center .
we assume that both the pscs and the mc are axisymmetric ( pscs are initially spherical but are likely to quickly flatten as time evolves ) .
the radial density profile of the mc is given by : @xmath10 where @xmath11 and @xmath12 are the cloud s central density and core radius , respectively .
the central density , @xmath11 , is given by : @xmath13 } , \label{eq2}\ ] ] where @xmath14 is the mass of the cloud and @xmath15 its radius .
the density profiles of pscs are assumed to follow the formula given by whitworth & ward - thompson ( 2001 ) : @xmath16^{2 } } , \label{eq3}\ ] ] where @xmath17 and @xmath18 are the central density and core radius of the psc , respectively .
the radius @xmath19 of the the psc depends both on its mass and on its position within the mc .
the dependence of @xmath19 on @xmath9 requires that the density at the edges of the psc equals the ambient cloud density , i.e. , @xmath20 .
this would result in smaller radii for pscs of a given mass when they are located in their inner parts of the cloud .
the density contrast between the edge of the psc and its center is given by @xmath21 depending on its position @xmath9 in the cloud , the radius of the psc of mass @xmath22 , @xmath19 , can be calculated as being @xmath23 , where : @xmath24-\frac{a(r)}{1+a(r)^{2 } } \right)^{-1/3 } , \label{eq5}\ ] ] and with @xmath25 . with our set of parameters , the quantity @xmath26 is always guaranteed to be positive .
the value @xmath27 can be considered as being the radius of the psc at the moment of its formation .
however , the radius of the psc will decrease as time advances due to gravitational contraction .
the psc contracts on a timescale , @xmath28 which is equal to a few times its free fall timescale , and can be parametrized as : @xmath29 where @xmath30 and @xmath31 is the radially averaged density of the psc of mass @xmath22 , located at position @xmath9 in the cloud , and which is calculated as being : @xmath32^{2 } } dr_{p } , \label{eq7}\ ] ] thus , the time evolution of the radius of the psc can be described by the following equation : @xmath33 once the instantaneous radius of a psc of mass @xmath22 , located at position @xmath9 form the cloud s center is defined , it becomes possible to calculate its cross section for collision with pscs of different masses .
the cross section for the collision of a psc of mass @xmath34 and radius @xmath35 with another of mass @xmath36 and radius @xmath37 and which accounts for the effect of gravitational focusing is given by : @xmath38^{2 } \nonumber \\
\times \left [ 1+\frac{2 g ( m_{i}+m_{j } ) } { 2 v^{2 } ( r_{p , i}(r , m_{i},t)+r_{p , j}(r , m_{j},t ) ) } \right ] .
\label{eq9}\end{aligned}\ ] ] elmegreen & shadmehri ( 2003 ) and shadmehri ( 2004 ) assumed that the collision velocity between pscs is equal to the virialized velocity dispersion inside the mc .
this might be a plausible hypothesis if mcs were indeed the dissipative structures of turbulence in the interstellar medium .
it is however unlikely to be the case .
numerical simulations ( e.g. , dib et al . 2007 ) show that clumps and cores in mcs are not in virial equilibrium . in this work
, we assume that the relative collision velocity between the pscs follows the local gas dynamics at their position in the cloud ( this remains a simplification as in reality pscs motions can be decoupled from that of the local ambient gas ) according to a larson type relation @xmath39 ( larson 1981 ; @xmath40 km s@xmath41 ) , with a lower limit being the local thermal sound speed , which is uniform across the isothermal mc .
as initial conditions for the pscs mass distribution at different cloud radii , we adopt distributions that are the result of the gravo - turbulent fragmentation of the cloud , following the formulation given in padoan et al .
( 1997 ) and padoan & nordlund ( 2002 ) . in these models ,
the probability function of the density field is well represented by a log - normal function : @xmath42 d~ln~x , \label{eq10}\ ] ] where @xmath43 is the number density normalized by the average number density , @xmath44 .
the standard deviation of the density distribution @xmath45 and the mean value @xmath46 are functions of the thermal rms mach number , @xmath47 : @xmath48 and @xmath49 .
padoan & nordlund ( 2002 ) suggest a value of @xmath50 . a second step in this approach is to determine the mass distribution of dense cores .
padoan & nordlund ( 2002 ) showed that by making the assumptions that : ( a ) the power spectrum of turbulence is a power law and , ( b ) the typical size of a dense core scales as the thickness of the postschock gas layer , the cores mass spectrum is given by : @xmath51 where @xmath52 is the exponent of the kinetic energy power spectrum , @xmath53 , and is related to the exponent @xmath54 of the size - velocity dispersion relation in the cloud with @xmath55 .
however , eq . [ eq11 ] can not be directly used to estimate the number of cores that are prone to star formation .
it must be multiplied by the local distribution of jeans masses . at constant temperature
, this distribution can be written as : @xmath56 dm_{j } , \label{eq12}\ ] ] where @xmath57 is the jeans mass at the mean density @xmath58 .
thus , eq . [ eq11 ] becomes , locally : @xmath59d~log~m .
\label{eq13}\ ] ] the local normalization coefficient @xmath60 is obtained by requiring that @xmath61 in the shell of width @xmath62 located at distance @xmath9 from the cloud s center .
then , the local distribution of cores at time @xmath63 , @xmath64 , is obtained by multiplying the local normalized function @xmath65 by the local rate of fragmentation such that : @xmath66 where @xmath67 is the average core mass in the local distribution and is calculated by @xmath68 and @xmath69 is a parameter smaller than unity which describes the local mass fraction of gas that is present in the dense pscs . in principle
, @xmath69 will have a radial and probably outwardly decreasing dependence .
for simplicity we shall assume @xmath69 to be a constant independent of radius .
as our comparisons with the observations will be focused on the inner parts of the protocluster cloud which will be transformed into a stellar cluster ( i.e. , the arches cluster ) , it is likely that these regions will be characterized by a uniform mass fraction of the dense gas .
[ fig1 ] displays the local mass spectrum of jeans unstable pscs in rings of width 0.025 pc , obtained with eq .
[ eq14 ] , located at different distances from the cloud s center ( top ) , as well as the cumulative number of pscs in each mass bin in regions of the protocluster cloud located between [ @xmath70 , [ @xmath71 , and [ @xmath72 .
with the initial conditions described in . 3
, we follow the time evolution of the pscs mass spectrum by solving the following integro - differential equation of @xmath73 : @xmath74 where the first and second terms in the right hand side of eq .
[ eq15 ] correspond to the rate of creation and destruction of a psc of mass @xmath75 , at location @xmath9 , respectively ( nakano 1966 ; shadmehri 2004 ) . in eq .
[ eq15 ] , @xmath76 , and @xmath77 is a coefficient which represents the coalescence efficiency , with @xmath78 .
this efficiency can be the result of various physical processes which can affect the coalescence of pscs , such as if the merger of cores occurs preferentially parallel or perpendicular to the local magnetic field lines , and is likely to have a radial dependence . for simplicity
, we shall assume that @xmath79 is independent of position . in order to evaluate the transition from pscs to stars ,
we compare , at each timestep , the local coalescence timescale to the local contraction timescale for pscs of a given mass .
the local coalescence timescale is @xmath80 where @xmath81 is the coalescence rate ( elmegreen & shadmehri 2003 ) : @xmath82 , \label{eq16}\ ] ] where @xmath83 is the number of mass bins , and @xmath84 is the volume of the shell of width @xmath62 located at distance @xmath9 from the mc s center .
the contraction timescale is given by eq .
[ eq8 ] . whenever the local contraction timescale is shorter than the local coalescence timescale , pscs
are collapsed into stars . when a psc collapses to form a star , we assume that a fraction of its mass is re - injected into the protocluster cloud in the form of an outflow
. we account for this mass loss in a purely phenomenological way by assuming that the mass of a star which is formed out of a psc of mass @xmath22 is given by m@xmath85=@xmath86 m@xmath87 , where @xmath88 .
matzner & mckee ( 2000 ) showed that @xmath86 can vary between @xmath89 for stars in the mass range @xmath90 @xmath1 .
there is no evidence so far , for or against , whether this result holds at higher masses .
however , the similarity between the imf and the dense cores mass function observed by alves et al .
( 2007 ) in the pipe nebula might be an indication of a constant @xmath86 across the mass spectrum ( i.e. , in their case it is @xmath91 ) . here
also , we shall assume that a similar fraction of the mass of a psc is lost in the outflow independent of its mass .
the algorithm was tested by performing runs with @xmath92 ( i.e. , no - coalescence ) and @xmath93 ( i.e. , inefficient coalescence ) and with the other parameters fixed at @xmath94 @xmath1 , @xmath95 pc , @xmath96 pc , @xmath97 @xmath98 , @xmath99 , @xmath100 , @xmath101 , and @xmath102 . as expected , for @xmath92 , the resulting stellar mass spectrum after the pscs collapse into stars is similar to the initial cumulative pscs spectrum , and only slightly modified for @xmath93 .
models were performed with permutations over the parameters @xmath79 and @xmath103 fixing the other parameters to the above stated values . it should be stressed at this stage that our semi - analytical modeling is not aimed at recovering the initial characteristics of the arches protocluster cloud , but rather at showing whether or not , the archer cluster imf can be generated by the coalescence of pscs and their subsequent collapse into stars .
[ fig2 ] displays the time evolution of the cumulative pscs populations in the region [ @xmath104=[0.2 - 0.4 ] pc , which corresponds to the annulus between @xmath105 1 - 2 core radii of the arches cluster for a model with @xmath106 and @xmath101 . in the initial stages , the most massive pscs , who have a larger cross section , coalesce faster than the less massive ones , essentially by capturing the numerous intermediate mass pscs and causing a rapid flattening of the spectrum at the high mass end .
by @xmath107 ( @xmath108 yr is the mc free fall timescale ) , a first generation of the smallest pscs collapses to form stars .
as time advances , more massive stars are formed in the shell ( massive cores collapse later because of their lower average density ) and in parallel the pscs population decreases . by @xmath109 the intermediate mass pscs which constitutes the largest mass reservoir for coalescence collapse into stars . at this time
, the turnover in the pscs mass spectrum is located at @xmath110 @xmath1 .
since the reservoir of intermediate mass objects is depleted , the remaining massive pscs coalesce at a slower pace before they collapse .
by @xmath111 , all pscs of different masses in the shell have collapsed and formed stars .
because of mass loss , the stellar imf is shifted to lower masses ( bump shifted to @xmath0 @xmath1 ) . in summary , the resulting imf is not very different from the pscs mass spectrum after the initial and rapid stage of strong coalescence until @xmath112 .
this is due to the fact that low and intermediate mass pscs collapse at early stages , thus depleting the reservoir of objects with which the massive pscs can continue to coalesce , in addition to their own contraction .
both effects reduce the massive pscs ulterior merger rate .
overall , the stellar mass spectrum is formed very quickly , on a timescale which is of the order of the contraction timescale of the most massive cores i.e. , @xmath113 . in fig .
[ fig2 ] , over - plotted to our result is the cumulative mass spectrum of the arches cluster in the annulus of [ 0.2 - 0.4 ] pc ( kim et al .
the coalescence - collapse model agrees better with the observations than models based on mass segregation by dynamical friction .
in particular , the bump at @xmath0 @xmath1 is reproduced .
a fit to the stellar spectrum yields slopes of @xmath114 and @xmath115 in the mass ranges of [ @xmath116 @xmath1 and @xmath117 @xmath1 , respectively , in very good agreement with observational values .
we also performed additional runs where the maximum mass in the psc spectrum was set to 250 @xmath1 ( instead of 100 @xmath1 ) . in this case , the resulting slope of the imf in the low and high mass regimes are shallower than the salpeter imf , yet shallower than those of the arches imf .
the reason is that pscs with masses larger than @xmath118 @xmath1 will form quickly from the coalescence of lower mass ones , and the number of pscs of masses @xmath119 @xmath1 will grow at an even faster pace as their cross sections are very large .
the mismatch in this case with the arches imf might be an indication that pscs with masses @xmath120 @xmath5 , if they form , might undergo a certain amount of sub - fragmentation .
in this work , we use semi - analytical modeling to study the evolution of the pre - stellar cores ( pscs ) mass spectrum and its transition to the stellar initial mass function ( imf ) at different locations in a molecular cloud ( mc ) under the assumption that the coalescence of pscs is important .
the aim is to reproduce the observed imf in the inner regions of dense stellar clusters such as the arches cluster ( kim et al .
the initial conditions for the local populations of pscs are those of jeans unstable cores resulting from the gravo - turbulent fragmentation of the mc .
pscs of a given mass are transformed into stars whenever their local rate of contraction is higher than their rate of coalescence . with appropriate , yet very realistic parameters , we are able to reproduce all of the observed characteristics of the imf of the arches cluster .
namely , the slopes at the high and low mass ends , and the peculiar bump observed at @xmath0 @xmath1 .
our results suggest that today s imf of the arches cluster is primordial .
this might be a common property of young and dense stellar clusters ( e.g. , chen et al .
another consequence of the coalescence - collapse model is that it might help explain the formation of intermediate - mass black holes ( @xmath121 @xmath1 ) in the central regions of dense stellar clusters , either by the direct gravitational collapse of massive pscs or by the runaway collisions of massive stars ( e.g. , bonnell et al .
1998 ; freitag et al . 2006 ) which would be fostered if the primordial imf is top - heavy .
we thank the anonymous referee for valuable comments and dongsu ryu , christopher matzner , carl jakob walcher , sungsoo kim , and zhi - yun li for useful discussions .
alves , j. , lombardi , m. , & lada , c. j. 2007 , a&a , 462 , l17 andr , p. , ward - thompson , d. , & barsony , m. 2000 , in protostars and planets iv , ed . v. mannings , a. p. boss , s. s. russel ( tuscon : university of arizona press ) , 59 bonnell , i. a. , bate , m. r. , & zinnecker , h. 1998 , mnras , 298 , 93 chabrier , g. 2003 , pasp , 115 , 763 chen , l. , de grijs , r. , & zhao , j. l. 2007 , aj , accepted , ( arxiv:0706.2723 ) dib , s. , kim , j. , vzquez - semadeni , e. , burkert , a. , & shadmehri , m. 2007 , apj , 661 , 262 eisenhauer , f. , quirrenbach , a. , zinnecker , h. , & genzel , r. 1998 , apj , 498 , 278 elmegreen , b. g. , & shadmehri , m. 2003 , mnras , 338 , 817 elmegreen , b. g. 2004 , mnras , 354 , 367 figer , d. f. , mclean , i. s. , & morris , m. 1999 , apj , 514 , 202 freitag , m. , atakan grkan , m. , & rasio , f. 2006 , mnras , 368 , 141 gouliermis , d. , keller , s. c. , kontizas , m. , kontizas , e. , & bellas - velidis , i. 2004 , a&a , 416 , 137 hillenbrand , l. a. , & hartmann , l. w. 1998 , apj , 492 , 540 kim , s. s. , figer , d. f. , kudritzki , r. p. , & najarro , f. 2006 , apj , 653 , l113 kroupa , p. 2002 , science , 295 , 82 larson , r. b. 1981 , mnras , 194 , 809 lin , d. n. c. , & murray , s. d. 2007 , apj , accepted , ( astro - ph/0703807 ) matzner , c. d. , & mckee , c. f. 2000 , apj , 545 , 364 miller , g. e. , & scalo , j. m. 1979 , apjs , 41 , 513 mouri , h. , & taniguchi , y. 2002 , apj , 580 , 844 nakano , t. 1966 , prog .
phys . , 36 , 515 padoan , p. , nordlund , . , & jones , b. j. t. 1997 , mnras , 288 , 145 padoan , p. , & nordlund , .
2002 , apj , 576 , 870 portegies zwart , s. , gaburov , e. , chen , h .- c . , &
atakan grkan , m. 2007 , mnras , accepted , ( astro - ph/0702693 ) rieke , g. h. , loken , k. , rieke , m. j. , & tamblyn , p. 1993 , apj , 412 , 99 salpeter , e. e. 1955 , apj , 121 , 161 shadmehri , m. 2004 , mnras , 354 , 373 silk , j. , & takahashi , t. 1979 , apj , 229 , 242 sternberg , a. 1998 , apj , 506 , 721 stolte , a. , grebel , e. k. , brandner , w. , & figer , d. f. 2002 , a&a , 394 , 459 stolte , a. , brandner , w. , grebel , e. k. , lenzen , r. , & lagrange , a .- m .
2005 , apj , 628 , l113 vesperini , e. , & heggie , d. c. 1997 , mnras , 289 , 898 whitworth , a. p. , & ward - thompson , d. 2001 , apj , 547 , 317 | we investigate the time evolution of the mass distribution of pre - stellar cores ( pscs ) and their transition to the initial stellar mass function ( imf ) in the central parts of a molecular cloud ( mc ) under the assumption that the coalescence of cores is important .
our aim is to explain the observed shallow imf in dense stellar clusters such as the arches cluster .
the initial distributions of pscs at various distances from the mc center are those of gravitationally unstable cores resulting from the gravo - turbulent fragmentation of the mc .
as time evolves , there is a competition between the pscs rates of coalescence and collapse . whenever the local rate of collapse is larger than the rate of coalescence in a given mass bin ,
cores are collapsed into stars . with appropriate parameters , we find that the coalescence - collapse model reproduces very well all the observed characteristics of the arches stellar cluster imf ; namely , the slopes at high and low mass ends and the peculiar bump observed at @xmath0 @xmath1 .
our results suggest that today s imf of the arches cluster is very similar to the primordial one and is prior to the dynamical effects of mass segregation becoming important .
[ firstpage ] galaxies : star clusters - galaxy : centre - turbulence - ism : clouds - open clusters and associations : individual : arches |
whitney - graustein theorem says that two regular closed curves on the euclidean plane @xmath2 are regularly homotopic if and only if they have the same ` rotation number ' ( _ a.k.a .
_ ` winding number')@xcite .
smale classified regular closed curves on an arbitrary surface up to regular homotopy ( fixing the base point and the base direction ) @xcite .
but he did not try to define the winding numbers ; he was more interested in the differences . winding
numbers of closed curves on surfaces were introduced by reinhart @xcite when the surface is orientable , and later extended by chillingworth in the non - orientable case @xcite .
these works are successful for surfaces with spherical or euclidean geometry , but they are too week to give a regular homotopy classification in the case of hyperbolic surfaces . the aim of this paper is to introduce two types of winding numbers of a regular closed curve on an _ aspherical _ surface , _ i.e. _ a surface with euclidean or hyperbolic geometry . let us review how the winding numbers were defined in earlier works . to a regular closed curve @xmath3\to { \mathbb r}^2 $ ]
, whitney associates an angle function @xmath4\to{\mathbb r}$ ] which measures the angle of the tangent vector from the positive direction of the @xmath5-axis , and defined the winding number to be @xmath6 . in the case of an arbitrary surface
, there is no preferred direction .
the basic idea of reinhart and chillingworth is to use a continuous vector field @xmath7 on the surface to measure how the tangent vector of the curve rotates with respect to @xmath7 .
this is fine in the euclidean case , because there are vector fileds with no singular points on them . in the other cases , we need to allow the vector field to have singular points .
then , when a regular homotopy hits a singular point of @xmath7 , the ` winding number ' jumps ! therefore , winding numbers are defined in some quotient @xmath8 of @xmath9 .
luckily , in the spherical case the @xmath10-winding number can be used to classify the regular closed curves . on the sphere
there are two regular homotopy classes ; on the projective plane there are two free homotopy classes of closed curves and in each of them there are two regular homotopy classes which can be distinguished by the @xmath10-winding numbers .
as long as we stick to vector fields on the surface , we will never get a winding number which is strong enough to classify the regular homotpy classes . since the spherical surfaces are quite exceptional among the surfaces , i propose to study only aspherical surfaces and use non - vanishing vector fields on the universal covers of the surfaces .
slightly more precisely , we consider a riemannian surface @xmath11 with a nice conformal ( or _ isogonal _ ) embedding of the universal cover @xmath12 into the euclidean plane @xmath2 .
such an embedding is called a _
tame isogonally euclidean structure _ of @xmath11 .
see 4.1 for the detail .
let @xmath3\to m$ ] be a regular closed curve on such a surface @xmath11 with base point @xmath13 .
fix a point @xmath14 in @xmath12 over @xmath15 , and take a lift @xmath16\to \widetilde m\subset { \mathbb r}^2 $ ] with @xmath17 .
as the vector field @xmath7 on @xmath12 , we use the vector @xmath18 at every point of @xmath12 .
note that winding numbers can be defined not only for regular closed curves but also for nonclosed regular curves on @xmath2 .
for example , a half circle has winding number @xmath19 .
such an extension is called the _ @xmath20-index _ of a regular curve @xmath21 , and is denoted @xmath22 : @xmath23 where @xmath24 is the angle function of @xmath21 .
we also define the _ @xmath25-index _ @xmath26 by @xmath27 as an element of @xmath28 .
these do not depend on the choice of the angle function .
suppose we locally rotate @xmath29 around the base point . if @xmath29 is orientation preserving , then the @xmath20-index @xmath22 of @xmath21 does not change . if @xmath29 is orientation reversing , then the @xmath25-index @xmath26 of @xmath21 does not change .
when @xmath30=1\in\pi_1(m , p)$ ] , we define its winding number @xmath31 to be @xmath22 . to define the winding number of a curve @xmath29 which represents a non - trivial element of @xmath32 , we need to correct the distorsion that comes from the difference of the riemannian metrics of @xmath12 and @xmath2 : choose the shortest geodesic @xmath33 on @xmath11 with base point @xmath15 which represents the same element of @xmath34 as @xmath29 , and take a lift @xmath35 of @xmath33 which starts at @xmath14 .
let @xmath36 be a suitably defined _ external angle _ ( 4.8 ) of @xmath33 at the vertex @xmath15 . if the curve @xmath29 on @xmath11 is orientation preserving , the _ winding number _
@xmath31 of @xmath29 _ with respect to _ @xmath14 is defined to be @xmath37 , which turns out to be an integer . if @xmath29 is orientation reversing , the _ winding number _
@xmath31 of @xmath29 _ with respect to _ @xmath14 is defined to be @xmath38 mod 2 , which is an integer modulo 2 . in 4 , we will prove the following : let @xmath39 and @xmath40 be regular closed curves on a surface @xmath11 with a tame isogonally euclidean structure @xmath41 .
suppose that they have the same base point @xmath15 and fix a lift @xmath14 of @xmath15 in @xmath12 .
then the following are equivalent .
\(1 ) @xmath39 and @xmath40 are regularly homotopic fixing the base point .
\(2 ) @xmath42=[\gamma_2]\in\pi_1(m , p)$ ] and @xmath43 .
we next study the behavior of @xmath31 when we change @xmath29 by a free regular homotopy .
for a regular closed curve @xmath29 on @xmath11 which is freely homotopic to a fixed closed curve @xmath44 , we define the _ free winding number @xmath45 of @xmath29 with respect to a lift @xmath46 of @xmath44_. the following is the main theorem of this paper .
let @xmath11 be a surface with a tame isogonally euclidean structure @xmath41 , and @xmath44 be a closed curve on @xmath11 .
suppose two regular closed curves @xmath39 and @xmath40 are freely homotopic to @xmath44 keeping the curve closed .
then the following are equivalent .
\(1 ) @xmath39 and @xmath40 are regularly homotopic .
\(2 ) @xmath47 .
the structure of this paper is as follows . in 2
, we develop a theory of @xmath20-indices and regular homotopies of possibly - non - closed regular curves on @xmath2 . note that smale claims that similar results can be obtained for nonclosed curves at the end of 7 of @xcite . instead of trying to follow smale s method , we will modify whitney s argument because we are only interested in the curves on @xmath2 .
more precisely , we will modify the proof of whitney - graustein theorem given in the textbook @xcite of umehara and yamada .
the main theorem of 2 is theorem 2.11 which says that two regular curves on the plane with the same end points and the same end directions are regularly homotopic fixing the end points and the end directions if and only if they have the same @xmath20-index .
theorem 2.11 will be proved in 3 .
our proof uses a sort of ` squeezing ' technique which has a flavor of the @xmath48-controlled topology of chapman - ferry - quinn . in 4
, we define the winding numbers @xmath31 and prove theorem 1.1 . in 5 , we define the free winding numbers @xmath49 and prove theorem 1.2 .
the methods are also involved with the ` squeezing ' technique . in 6
, we consider the eulidean examples .
this section is based on the discussions with the following students of the 2015/2016 senior seminar in geometry : c. hanaoka , s. sakaguchi , t. teraoka , k. tsuro , s. uehara , m. yatsuka and t. yokota . in 7 , we give a combinatorial formula for the @xmath20-indices of a nonclosed curve on the euclidean plane .
we use whitney s formula @xcite for the winding numbers of closed plane curves .
a ` curve on the plane ' means a parametrized curve @xmath3\to{\mathbb r}^2 $ ] in this article .
a @xmath50-curve @xmath3\to { \mathbb r}^2 $ ] on the plane is said to be _ regular _ if the derivative @xmath51 is non - zero for every @xmath52 $ ] .
the derivative is also called the _ velocity _ of @xmath29 .
the unit vector @xmath53 is called the _ direction _ of @xmath29 .
the unit vectors @xmath54 and @xmath55 are called the _ initial direction _ and the _ terminal direction _ of @xmath29 , respectively .
they are also called the _ end directions_. for any @xmath56)$ ] , the following identity holds : @xmath57 thus @xmath58 determines the curve up to parallel translation . if @xmath3\to { \mathbb r}^2 $ ] is a regular curve and @xmath59\to [ c , d]$ ] is an orientation - preserving diffeomorphism ( _ i.e. _ a differentiable homeomorphism such that @xmath60 for all @xmath61 $ ] ) , then @xmath62\to{\mathbb r}^2 $ ] is also a regular curve with the same image .
this is called a _ reparametrization _ of a regular curve .
we identify @xmath29 and @xmath63 .
in fact , it is customary to write @xmath64 for example , let @xmath65 be the _ arc length parameter _ of a regular curve @xmath3\to{\mathbb r}^2 $ ] of length @xmath66 ; _ i.e. _ @xmath67 then @xmath68 is always positive , and @xmath69 defines a diffeomorphism @xmath70\to[0,l]$ ] with inverse @xmath71 . as i mentioned above , @xmath72 is denoted @xmath73 , and @xmath74 is a unit vector . here
@xmath75 and @xmath76 actually mean @xmath77 and @xmath78 , respectively .
a continuous function @xmath4\to { \mathbb r}$ ] is an _ angle function _ for a regular curve @xmath29 if @xmath79 for each @xmath52 $ ] .
\(1 ) any regular curve has an angle function .
if @xmath80 and @xmath81 are angle functions for @xmath29 , then there is an integer @xmath82 such that the following identity holds : @xmath83 \(2 ) when the angle function @xmath80 is differentiable with respect to the arc length parameter @xmath65 , then @xmath84 is the curvature of the curve .
the _ @xmath20-index _ @xmath85 and the _ @xmath25-index
_ @xmath86 of a regular curve @xmath3\to { \mathbb r}^2 $ ] are defined as follows : @xmath87 where @xmath80 is any angle function for @xmath29 .
they are independent of the choice of the angle function @xmath80 ; for example , if we add @xmath88 to the function @xmath80 , then @xmath89 does not change and @xmath90 increases by @xmath91 .
the @xmath20-index is an obvious extension of the classical winding numbers of regular closed curves on the plane to possibly non - closed regular curves .
let us use the following notation : for a real number @xmath92 , the unit vector @xmath93 will be donoted @xmath94 .
this number @xmath92 will be called the _
angle _ of the direction @xmath94 .
of course the angle is well - defined only modulo @xmath88 .
the angle of a unit vector @xmath95 is denoted @xmath96 . if @xmath92 and @xmath97 are the angles for the directions @xmath54 and @xmath55 , respectively , then @xmath98 in particular
, the @xmath20-index @xmath85 is an integer if @xmath99 , and the @xmath25-index @xmath86 is an integer modulo 2 if @xmath54 and @xmath55 are reflections of each other with respect to a horizontal line .
since @xmath100 and @xmath101 for some integers @xmath82 and @xmath66 , this is obvious .
a homotopy @xmath102\to{\mathbb r}^2\}_{0\leq t \leq 1}$ ] between two @xmath50-curves @xmath103\to{\mathbb r}^2 $ ] ( @xmath104 ) is said to be _ regular _ if 1 .
@xmath105 is a regular curve for each @xmath106 ( @xmath107 ) , and 2 .
the velocity @xmath108 is continuous with respect to @xmath106 .
if @xmath109 is a regular homotopy , @xmath110 and @xmath111 are continuous functions of @xmath106 . by choosing a suitable angle function @xmath112 for @xmath105
, we obtain a homotopy between @xmath113 and @xmath114 .
let @xmath115\to{\mathbb r}^2 $ ] be two regular curves such that @xmath116
. then @xmath117 implies @xmath118 .
this immediately follows from the next lemma which can be easily checked . if @xmath5 , @xmath119 , @xmath120 , @xmath121 are real numbers satisfying @xmath122 and @xmath123 , then @xmath124 .
the following is the main technical tool .
let @xmath15 , @xmath125 be points in @xmath2 , and @xmath94 , @xmath126 be two unit vectors in @xmath2 .
assume that two regular curves @xmath39 , @xmath127\to{\mathbb r}^2 $ ] and their directions @xmath128 , @xmath129 satisfy the following boundary conditions : @xmath130 and let @xmath131\to { \mathbb r}$ ] be angle functions for @xmath132 @xmath133 .
there exists a regular homotopy between @xmath39 and @xmath40 that fixes the end points and the end directions if and only if @xmath134 .
since this is not explicitly stated in @xcite , we will give an elementary proof in the next section .
the ` only if ' part is easy .
let @xmath105 be a regular homotopy between @xmath39 and @xmath40 satisfying the required conditions . then @xmath110 s are elements of the discrete subset @xmath135 of @xmath136 .
therefore , by 2.8 , @xmath110 is a constant function of @xmath106 .
let us first assume that @xmath137 .
the @xmath138 case will be handled at the end of this section . in the @xmath137 case , the ` if ' part is a consequence of the following special case .
suppose two regular curves @xmath139\to{\mathbb r}^2 $ ] satisfy the following conditions : * @xmath140 , @xmath141 @xmath133 , where @xmath142 is a positive number , * the direction @xmath143 of @xmath132 satisfies @xmath144 for @xmath104 , and * @xmath145 .
then there exists a regular homotopy between @xmath39 and @xmath40 that fixes the end points and the end directions .
we will assume this lemma and give a proof of the theorem in the @xmath137 case .
the lemma itself will be proved after that .
let @xmath39 and @xmath40 be regular curves satisfying the conditions posited in the theorem , and let @xmath146 and @xmath147 be their angle functions .
we may assume that @xmath148 for some integer @xmath82 .
we modify these curves in three steps so that the lemma above can be applied to the resulted two curves .
the first step is a diffeotopy from the identity to a parallel translation that sends @xmath15 to the origin : @xmath149 @xmath150 , and the curves @xmath151 s are curves from the origin to @xmath152 .
this does not change their angle functions , and their @xmath20-indices are the same .
the next step is a rotation around the origin . for a point @xmath153 and a real number @xmath154 , the rotation of angle @xmath154 around @xmath5
will be denoted @xmath155 .
choose an angle @xmath154 so that the rotation @xmath156 moves the point @xmath152 to @xmath157 , where @xmath158 , and set @xmath159 then @xmath160 s are curves from the origin to @xmath157 , and their angle functions are given by @xmath161 , @xmath104 ; therefore , their @xmath20-indices are the same .
note that their initial directions are equal to @xmath162 , and their terminal directions are equal to @xmath163 .
the third step is a diffeotopy that locally twists a neighborhood of 0 and a neighborhood of @xmath164 .
first choose two positive numbers @xmath165 , @xmath166 so that @xmath167 and prepare a smooth function @xmath168\to { \mathbb r}$ ] such that * @xmath169 if @xmath170 , * @xmath171 if @xmath172 , and * @xmath173 is decreasing on the interval @xmath174 $ ] .
set @xmath175 for any @xmath106 ( @xmath176 ) , all the points moved by @xmath177 are in the @xmath166-neighborhood of @xmath178 ; therefore @xmath177 is a well - defined diffeomorphisms .
now set @xmath179 , and define @xmath180 to be @xmath181 ( @xmath104 ) .
these are curves from 0 to @xmath164 , and their angle functions , denoted @xmath182 , have the following values at the end points : @xmath183 therefore , @xmath184 . also note that the directions of the curves @xmath185 and @xmath186 at the end points are all @xmath18 .
thus we can apply the lemma above to conclude that there is a regular homotopy @xmath187\to{\mathbb r}^2 $ ] between @xmath185 and @xmath186 .
now consider the regular homotopy @xmath188 .
when @xmath189 and @xmath190 , @xmath191 therefore it is a desired regular homotopy between @xmath39 and @xmath40 .
now let us prove lemma 3.1 .
first of all , we may assume that @xmath192 .
there is no loss of generality .
next , we may assume that the curves @xmath39 and @xmath40 have the same length @xmath66 .
if @xmath39 is shorter than @xmath40 , for example , then use a ` finger move ' on @xmath39 , as illustrated below , to make it longer .
this does not change the @xmath20-index of @xmath39 . then reparametrize the curves by the arc length parameter @xmath65 ( @xmath193 ) . obviously , an inequality @xmath194 holds .
now we squeeze the curves in the direction of the origin by regular homotopies , and add a straight tail at the terminal point .
let us temporarily drop the subscript @xmath25 and let @xmath29 represent both of @xmath132 s ( @xmath104 ) . for every @xmath106 satisfying @xmath195 , we define a curve @xmath196 as follows : the first part of the curve is @xmath197 of length @xmath198 , which starts from 0 and ends at @xmath199 , and the second part is a straight horizontal line segment of length @xmath200 which starts at @xmath199 and ends at @xmath201 .
the length of @xmath196 is @xmath202 , so we parametrize @xmath196 by @xmath65 ( @xmath193 ) so that it has the constant speed @xmath203 .
let us define a ( small ) positive number @xmath48 by @xmath204 then @xmath196 ( @xmath205 ) can be thought of as a regular homotopy between @xmath206 and @xmath29 , after a reparametrization of @xmath106 . throughout the regular homotopy
, the end points and the end directions are fixed .
let us denote the curve @xmath206 for @xmath207 by @xmath208 ( @xmath104 ) .
they both have the same length @xmath209 we reparametrize these by arc length parameter @xmath65 ( @xmath210 ) . now
the speed is 1 , and the time it takes a particle to move along the curves from the origin to @xmath211 is @xmath212 the picture below shows the relations of various numbers . the curves @xmath213 and @xmath214 have the same @xmath20-index @xmath215 .
let @xmath216 and @xmath217 be the angle functions of @xmath213 and @xmath214 such that @xmath218 note that we actually have @xmath219 define a homotopy @xmath112 ( @xmath176 ) between @xmath216 and @xmath217 by @xmath220 and define a vector function @xmath221 by : @xmath222 note that , for every @xmath106 , we have @xmath223 for any @xmath106 , the integral @xmath224 defines a regular curve which starts from the origin with velocity @xmath18 and reaches its terminal point @xmath225 with velocity @xmath18 .
these determine a regular homotopy between @xmath213 and @xmath214 , but unfortunately the terminal point may not be fixed .
let us denote the difference of the terminal point and @xmath201 by @xmath226 : @xmath227 we can estimate the length of this vector as follows : @xmath228 recall that we are assuming that @xmath192 and note that @xmath229 .
let @xmath230 \to{\mathbb r}$ ] be a smooth function satisfying the following conditions : * @xmath231 ( @xmath210 ) * @xmath232 if @xmath233 , @xmath234 , and * @xmath235 .
we modify @xmath236 and define a new function @xmath237 as follows : @xmath238 note that it is never 0 because @xmath239 and @xmath240 .
also note that @xmath241 for @xmath242 and @xmath234 .
in particular , we have @xmath243 now define a homotopy @xmath105 between @xmath213 and @xmath214 by @xmath244 the terminal point of the curve @xmath105 can be calculated as follows : @xmath245 thus @xmath105 is a regular homotopy between @xmath213 and @xmath214 which fixes the end points and the end directions .
this completes the proof of lemma 3.1 , and also the proof of theorem 2.11 under the assumption @xmath137 .
now let us consider the @xmath138 case . by a small regular homotopy fixing the end points and end directions
, we may assume that there exists an @xmath246 such that * @xmath247 if @xmath248 or @xmath249 , and * @xmath250 .
then we can apply the @xmath137 case to @xmath251 $ ] to get a desired regular homotopy between @xmath39 and @xmath40 .
this completes the proof of 2.11 .
in this section we consider regular closed curves on a smooth surface @xmath11 with a good euclidean or hyperbolic geometry , and study the regular homotopy classification of regular closed curves fixing the base point .
we use a winding number , denoted @xmath31 , which was already briefly described in the introduction .
we always assume that the surface @xmath11 has no boundary . if a surface has a non - empty boundary and if a curve or a homotopy touches the boundary , we can always push it into the interior of the surface using the collar structure near the boundary .
so the regular homotopy classification on the surface and the regular homotopy classification on its interior are the same .
we denote the universal cover of @xmath11 by @xmath12 and denote the projection from @xmath12 to @xmath11 by @xmath252 .
we equip @xmath12 with the riemannian metric induced from that of @xmath11 .
an _ isogonally euclidean structure _ on a riemannian surface @xmath11 is an embedding of @xmath12 into the euclidean plane @xmath2 which preserves the magnitudes of local angles .
we regard @xmath12 as a subset of @xmath2 using the embedding .
an isogonally euclidean structure is said to be _ tame _ if the following conditions hold : ( 1 ) any two points of @xmath12 can be joined by a unique shortest geodesic in @xmath12 , and ( 2 ) for any compact subset @xmath253 of the interior of @xmath12 , there is a regular map @xmath254 such that the restriction of @xmath255 to @xmath253 is the identity map .
the following riemannian surfaces have a tame isogonally euclidean structure : 1 . the euclidean space @xmath2 , 2 . the cylinder @xmath256\times{\mathbb r}/(0,y)\sim ( 1,y)$ ] ( @xmath257 ) , 3 .
the open annulus @xmath258\times(-\frac{1}{2},\frac{1}{2})/(0,y)\sim ( 1,y)$ ] ( @xmath259 ) , 4 .
the twisted cylinder @xmath260\times{\mathbb r}/(0,y)\sim ( 1,-y)$ ] ( @xmath261 ) , 5 .
the open mbius band @xmath262\times(-\frac{1}{2},\frac{1}{2})/(0,y)\sim ( 1,-y)$ ] ( @xmath263 ) , 6 . the torus @xmath264\times [ -\frac{1}{2},\frac{1}{2}]/(0,y)\sim(1,y ) , ~(x,-\frac{1}{2})\sim(x,\frac{1}{2})$ ] ( @xmath265 ) , 7 .
the klein bottle @xmath266\times [ -\frac{1}{2},\frac{1}{2}]/(0,y)\sim(1,1-y),~(x,-\frac{1}{2})\sim(x,\frac{1}{2})$ ] ( @xmath267 ) , 8 . any closed hyperbolic surface @xmath11 ( @xmath268 ) , 9 . any complete hyperbolic surface @xmath11 ( @xmath268 ) , where @xmath269 denotes either the poincar disc model or the upper half plane model of the hyperbolic plane .
( 2 ) and ( 3 ) are topologically the same , and so are ( 4 ) and ( 5 ) . in the first seven examples , the inclusion maps @xmath270
are actually isometries , and these are called the _ standard structures _ on @xmath2 , @xmath271 , @xmath272 , @xmath273 , @xmath274 , @xmath275 , and @xmath276 .
let @xmath15 be a point of a surface @xmath11 with an isogonally euclidean structure @xmath41 , and fix a preferred lift @xmath277 of @xmath15 .
let @xmath278\to m$ ] be a closed curve based at @xmath15 , and let @xmath279 be a lift of @xmath280 such that @xmath281 .
this determines a deck transformation @xmath282 be a point of @xmath12 .
connect @xmath283 to @xmath14 by an arc @xmath284\to \widetilde m$ ] .
take a lift @xmath285 of the arc @xmath286 such that @xmath287 .
then @xmath288 is equal to @xmath289 . ] which sends @xmath290 to @xmath291 . a closed curve @xmath278\to m$ ]
is said to be _ orientation preserving _ or _ orientation reversing _ according as @xmath292 preserves the orientation of @xmath12 or not .
this does not depend on the choice of @xmath279 . here
@xmath12 is given the orientation determined by the standard orientation of @xmath2 .
the _ sign _
@xmath293 of @xmath280 is defined by : @xmath294 the points @xmath295 and @xmath291 are said to have the _ same _ ( resp .
_ opposite _ ) _ orientation _ if @xmath280 is orientation preserving ( resp .
reversing ) , and we write @xmath296 ( resp .
@xmath297 ) . now let @xmath11 be a surface with a tame isogonally euclidean structure @xmath41 and let us consider a smooth curve @xmath3\to m$ ] on @xmath11 .
the velocity vector @xmath298 will be denoted @xmath299 as in the plane curve case .
if @xmath299 is never zero , then we say @xmath29 is _ regular _ , and @xmath300 will be denoted @xmath301 .
a curve @xmath3\to m$ ] is said to be a _ regular closed curve _ if it is reguar , closed ( _ i.e. _ @xmath302 ) , and @xmath99 . the point @xmath302 is called the _ base point _ of @xmath29 , and @xmath99 is called the _ base direction _ of @xmath29 .
_ regular homotopies _ of regular curves are defined in the obvious way .
when we consider a regular homotopy @xmath105 ( @xmath176 ) between regular closed curves , we also require that each @xmath105 is a _ regular closed curve _ in the sense above . as before ,
let us choose a point @xmath13 and fix a point @xmath277 over @xmath15 .
the first step of the regular homotopy classification is the following .
let @xmath39 and @xmath40 be regular closed curves on a surface @xmath11 with a tame isogonally euclidean structure @xmath41 .
suppose they have the same base point @xmath15 and the same base direction .
let @xmath303 and @xmath304 be the lifts of @xmath39 and @xmath40 with initial point @xmath14 .
then the following are equivalent . * @xmath39 and @xmath40 are regularly homotopic fixing the base point and the base direction .
* @xmath42=[\gamma_2]\in\pi_1(m , p)$ ] and @xmath305 .
( 1)@xmath306(2 ) : since a regular homotopy is a homotopy , @xmath42=[\gamma_2]\in\pi_1(m , p)$ ] is obvious .
since the regular homotopy lifts to a homotopy of @xmath303 and @xmath304 which fixes the ends and the end directions , @xmath305 follows from 2.11 .
( 2)@xmath306(1 ) by the condition @xmath42=[\gamma_2]\in \pi_1(m , p)$ ] , the terminal points of @xmath303 and @xmath304 are the same . since @xmath132 s have the same base direction , @xmath307 s have the same end directions .
now we obtain a desired regular homotopy between @xmath307 s in @xmath2 by 2.11 . by the tameness of the structure
, we can squeeze the regular homotopy into @xmath12 fixing @xmath307 s .
now project this regular homotopy down to @xmath11 to get a desired regular homotopy on @xmath11 .
we next study regular homotopies which fix the base point but not necessarily fix the base direction .
the following is a key fact about @xmath20-indices and @xmath25-indices introduced in 1 .
let @xmath3\to m$ ] be a regular closed curve based at @xmath15 , and @xmath21 be a lift of @xmath29 whose initial point is a lift @xmath14 of @xmath15 .
* when @xmath29 is orientation preserving , a local rotation of @xmath29 around the base point does not change the @xmath20-index @xmath22 of @xmath21 .
* when @xmath29 is orientation reversing , a local rotation of @xmath29 around the base point does not change the @xmath25-index @xmath26 of @xmath21 .
suppose we locally rotate @xmath29 around the base point so that @xmath21 rotates aroud @xmath308 by an angle @xmath92 .
then the other end of @xmath21 rotates around @xmath309 by an angle @xmath310 , because the deck transformation @xmath106 of @xmath12 which sends @xmath308 to @xmath309 is isogonal and @xmath106 is orientation preserving ( resp .
orientation reversing ) if @xmath29 is orientation preserving ( resp . orientation reversing ) .
suppose @xmath30=1\in\pi_1(m , p)$ ] .
the _ winding number _
@xmath31 of @xmath29 _ with respect to _ @xmath14 is defined to be @xmath311 , where @xmath21 is the lift of @xmath29 in @xmath12 whose intial point is @xmath14 . to define @xmath31 when @xmath312\ne 1\in\pi_1(m , p)$ ] , we need some preparations . by the assumption that @xmath29 is not null - homotopic , @xmath313 and @xmath309 are distinct points .
we first pick up the unique geodesic @xmath314\to\widetilde m$ ] connecting these two points in @xmath12 with the riemannian metric induced from that of @xmath11 .
this depends only on the homotopy class @xmath280 of @xmath29 in @xmath32 and on the choice of @xmath14 .
the geodesic @xmath315 on @xmath11 will be denoted @xmath33 , where @xmath316 is the covering projection .
this @xmath33 is the shortest geodesic based at @xmath15 that represents @xmath280 in @xmath32 , and can be regarded as the simplest closed curve on @xmath11 that represents @xmath280 .
so we would like to compare @xmath29 with @xmath33 .
note that @xmath33 may have a corner at the base point @xmath15 ; so , we need to take the external angle of @xmath33 at the vertex @xmath15 into consideration .
fortunately , the internal angle of @xmath33 at @xmath15 is not 0 , because a point and a direction determines a geodesic which passes the given point and has the given direction there ; therefore , the external angle can not be equal to @xmath317 .
since we are not assuming that @xmath11 is orientable , we fix the orientation convention as follows .
let @xmath35 be as above , and let @xmath318\to\widetilde m$ ] be a lift of @xmath33 whose initial point @xmath319 is equal to @xmath320 .
the _ external angle @xmath321 of @xmath33 with respect to _ @xmath14 is defined to be the unique number @xmath80 satisfying @xmath322 suppose @xmath30\ne 1\in\pi_1(m , p)$ ] .
the _ winding number _
@xmath31 of @xmath29 _ with respect to _ @xmath14 is defined as follows : * if @xmath29 is orientation preserving , then @xmath323 , and , * if @xmath29 is orientation reversing , then @xmath324 .
let @xmath29 be a regular closed curve based at @xmath15 .
* if @xmath29 is orientation preserving , then @xmath31 is an integer . *
if @xmath29 is orientation reversing , then @xmath31 is an integer modulo 2 .
the case when @xmath29 is null - homotopic is already done ; so we consider the non - null - hopmotopic case .
first we describe a method to convert @xmath33 into a regular closed curve which is regularly homotopic to @xmath29 fixing the base point and the base direction .
see the left picture below of @xmath21 and @xmath35 .
let @xmath92 ( resp .
@xmath97 ) be the angle between the direction of @xmath21 and @xmath35 at the initial point ( resp .
the terminal point ) .
make a local rotation of @xmath35 around the initial point by the angle @xmath92 , and a local rotation of @xmath35 around the terminal point by the angle @xmath97 .
now @xmath35 and @xmath21 have the same ends and the same end directions . if necessary ,
add @xmath215 small twists so that the result @xmath325 has the same @xmath20-index as @xmath21 , for some @xmath326 .
4.10 follows from the following propositions 4.11 and 4.12 .
@xmath327 , and @xmath328 .
if we choose an angle function @xmath329 which coincides with @xmath330 at a point away from the local twist near @xmath14 and before the @xmath215 twists , we have the following identities : @xmath331 subtracting ( 4.2 ) from ( 4.1 ) , we obtain @xmath332 and @xmath333 . by the choice of @xmath215 , @xmath22 and @xmath334 are the same ; therefore , we have @xmath327 . adding ( 4.2 ) and ( 4.1 ) , we obtain @xmath335 and @xmath336 . since @xmath337 and the directions @xmath338 and @xmath339 are the same , we have @xmath340 mod 2 by 2.9 .
therefore , we have @xmath328 . @xmath341 .
the external angle of @xmath33 can be computed by looking at the picture around @xmath309 .
the left picture below describes the relevant directions and the rotations in the case when @xmath29 is orientation preserving , and the right picture shows the case when @xmath29 is orientation reversing , where @xmath342 denotes the lift of @xmath29 whose initial point is @xmath309 .
the external angle can be computed as follows : @xmath343 * proof of theorem 1.1 .
* \(1 ) @xmath306 ( 2 ) : @xmath42=[\gamma_2]\in\pi_1(m , p)$ ] is obvious .
let @xmath307 ( @xmath104 ) be the lift of @xmath132 whose initial point is @xmath14 .
then @xmath307 s are regularly homotopic fixing the base point .
by such a regular homotopy , the @xmath20-indices and @xmath25-indices change continuously , and @xmath33 is unchanged ; therefore , @xmath31 changes continuously . but the value set is discrete , so it does not change .
\(2 ) @xmath306 ( 1 ) : let @xmath307 ( @xmath104 ) be the lift of @xmath132 whose initial point is @xmath14 .
first assume that @xmath39 and @xmath40 are null - homotopic .
then @xmath307 s are regular closed curves based at @xmath14 . by a local rotation of @xmath40 around the base point
, we may assume that @xmath132 s have the same direction at the base point .
note that a local rotation is a regular homotopy , and that , by 4.6 , it does not change @xmath344 . by 2.11
, there is a regular homotopy between @xmath303 and @xmath304 in @xmath2 . as in the proof of 4.5
, we can use the tameness condition to obtain a desired regular homotopy between @xmath39 and @xmath40 .
next assume that @xmath39 and @xmath40 are not null - homotopic . by the condition @xmath42=[\gamma_2]\in \pi_1(m , p)$ ] , the terminal points of @xmath303 and @xmath304 are the same .
a local rotation of @xmath132 around @xmath15 is a regular homotopy , so it does not change the situation , and we may assume that the initial directions of @xmath303 and @xmath304 are both @xmath18 , and the ternminal directions of @xmath303 and @xmath304 are the same . let @xmath345 denote the common terminal point of @xmath307 s .
if @xmath346 $ ] is orientation preserving , then @xmath43 implies that @xmath347 and the result follows from 2.11 and the tameness condition .
suppose @xmath346 $ ] is orientation reversing ; choose the angle function @xmath348 of @xmath307 satisfying @xmath349 ( @xmath104 )
. then we have @xmath350 and @xmath43 implies that there exists an integer @xmath82 such that @xmath351 now let us locally rotate @xmath40 around @xmath15 by an angle of magnitude @xmath352 to get a new curve @xmath353 .
this induces local rotations of @xmath304 around @xmath14 and @xmath345 by angle of magnitude @xmath352 , but the sign of the angles are opposite .
we choose the rotation so that the new angle function @xmath354 satisfies @xmath355 then @xmath356 now we have a regular homotopy between @xmath39 and @xmath353 in @xmath2 by 2.11 . an argument which uses the tameness condition finishes the proof of theorem 1.1 .
@xmath357 we close this section by proving the following formula .
let @xmath14 and @xmath358 be two lifts of @xmath15
. then @xmath359 .
let @xmath21 and @xmath342 be the lifts of @xmath29 whose initial points are @xmath14 and @xmath358 , respectiely .
note that the deck transformation @xmath106 that sends @xmath14 to @xmath358 maps @xmath21 to @xmath342 isometrically with respect to the riemannian metric coming from that of @xmath11 , but not isometrically with respect to the euclidean metric .
also recall that @xmath360 is @xmath361 if the deck transformation @xmath106 is orientation preserving and it is @xmath297 if @xmath106 is orientation reversing .
first assume that @xmath29 is null - homotopic .
the two curves @xmath21 and @xmath342 are both regular closed curves .
if we squeeze @xmath21 into a small neighborhood of @xmath14 by a regular homotopy fixing the base and the base direction , then @xmath29 and @xmath342 also deform into small neighborhoods of the bases in similar ways .
this does not change @xmath22 and @xmath362 .
if we keep squeezing @xmath21 smaller and smaller , the geometry coming from @xmath12 around @xmath21 and @xmath342 becomes almost euclidean , and the deck transformation @xmath106 sends @xmath21 to @xmath342 almost isometrically with respect to the euclidean metric .
so , if @xmath106 preserves the orientation , @xmath22 must be equal to @xmath362 , and , if @xmath106 reverses the orientation , then @xmath22 must be equal to @xmath363 .
next assume that @xmath29 is not null - homotopic .
let @xmath35 be the shortest geodesic in @xmath12 that connects @xmath308 and @xmath309 .
then @xmath364 is the shortest geodesic in @xmath12 connecting @xmath365 and @xmath366 .
let us first consider the case when @xmath29 is orientation preserving .
we prove the following two identities which confirm the proposition .
@xmath367 recall from 4.11 that @xmath368 can be computed from the local data @xmath92 , @xmath97 , and @xmath215 . if @xmath106 is orientation preserving ( _ i.e. _ @xmath369 ) , then @xmath370 can be computed using the same @xmath92 , @xmath97 , and @xmath215 . if @xmath106 is orientation reversing ( _ i.e. _ @xmath371 ) , then @xmath370 can be computed using @xmath372 , @xmath373 , and @xmath374 .
thus we have the first identity . the external angles @xmath321 and @xmath375 can be also computed by the local data around @xmath309 and @xmath366
; therefore , we have the second identity , and the proposition is proved when @xmath29 is orientation preserving . when @xmath29 is orientation reversing , we study @xmath25-indices instead of @xmath20-indices .
the argument is exactly the same .
if either @xmath11 is orientable or @xmath11 is non - orientable and @xmath29 is orientation preserving , then @xmath31 does not depend on the choice of @xmath14 . in this case
, we can omit @xmath14 from the notation .
in this section , we study the free regular homotopy classification of regular closed curves on surfaces with a tame isogonally euclidean structure . we first fix a closed curve @xmath44 in @xmath11 based at @xmath376 , and , for a regular closed curve @xmath29 which is freely homotopic to @xmath44 as closed curves , we define the _
free winding number @xmath45 of @xmath29 with respect to _ a lift @xmath46 of @xmath44 , which is used for the free regular homotopy classification 1.2 . in this section
we sometimes need to compose consecutive curves / homotopies .
usually this needs reparametrization of the curves / homotopies , but we do not mention such reparametrizations .
let us look at the effect of a regular homotopy of regular closed curves on the winding number @xmath31 .
suppose that @xmath377 @xmath378 is a regular homotopy between regular closed curves @xmath39 and @xmath40 on @xmath11 , and let @xmath379 be a lift of the homotopy @xmath377 in @xmath12 .
let @xmath380 and @xmath381 be the initial points of @xmath303 and @xmath304 , respectively
. then the identity @xmath382 holds .
first let us consider the case when @xmath132 s are null - homotopic . in this case
, @xmath31 is the integer @xmath22 , where @xmath21 is a lift of @xmath29 with initial point @xmath14 .
the regular homotopy of @xmath21 changes @xmath383 continuously and the set of integers are discrete ; therefore , it must be constant .
the case when @xmath132 s are not null - homotopic can be handled similarly , because @xmath368 , @xmath384 , and the external angle @xmath321 change continuously .
let @xmath29 be a regular closed curve with base point @xmath15 , and @xmath385 is a regular curve with initial point @xmath15 such that the initial direction of @xmath33 and the base direction of @xmath29 are linearly independent .
a finger move of the base point of @xmath29 along the curve @xmath33 is a typical example of a regular homotopy . when the curve @xmath33 is closed , the finger move of this type is homotopically equivalent to taking the conjugate of the class of @xmath29 by the element represented by @xmath33 .
a closed curve @xmath29 of a non - orientable surface @xmath11 based at @xmath15 is said to be _
reversible _ if there is an orientation reversing closed curve @xmath280 based at @xmath15 such that the `` conjugate '' @xmath386 is homotopic to @xmath29 fixing the base point , where @xmath387 is the inverse of @xmath280 . if a closed curve is not reversible , then it is said to be _ non - reversible_. reversibility depends only on the free homotopy class of closed curves .
suppose there is a free homotopy @xmath388 between two closed curves @xmath29 and @xmath389 which keeps the curve closed , and let @xmath390 be the induced curve from the base point @xmath15 of @xmath29 to the base point @xmath125 of @xmath389 .
assume that @xmath29 is reversible and let @xmath280 be an orientation reversing closed curve based at @xmath164 such that there is a homotopy @xmath253 between @xmath29 and @xmath386 which keeps the base point fixed . consider the closed curve @xmath391 based at @xmath125 .
let @xmath14 be a lift of @xmath15 in @xmath12 , and let @xmath392 denote a lift of @xmath390 with initial point @xmath14 .
then it can be easily checked that the deck transformation @xmath292 is the same as the deck transformation @xmath393 ; therefore , @xmath391 is orientation reversing .
now we can show that @xmath389 is also reversible , because the conjugate @xmath394 of @xmath389 by the orientation reversing closed curve @xmath391 is homotopic to @xmath389 fixing the base point as in the picture below . here
the top and bottom rectangles on the left are filled with the free homotopy @xmath388 , and the middle rectangle is filled with a homotopy induced from @xmath253 .
let @xmath11 be a surface with a tame isogonally euclidean sturcture and @xmath29 be a regular closed curve on @xmath11 based at @xmath13 .
free winding number _
@xmath45 of @xmath29 is defined as follows .
\(1 ) if @xmath11 is orientable , then @xmath395 .
\(2 ) if @xmath11 is non - orientable and @xmath29 is orientation reversing , then @xmath396 .
\(3 ) if @xmath11 is non - orientable , and @xmath29 is orientation preserving and reversible , then @xmath397 .
here @xmath14 is any lift of @xmath15 . by 4.13 ,
the free winding number is well - defined in these cases , and , by 5.1 , it is a regular homotopy invariant of @xmath29 .
note that it does not depend on the choice of @xmath44 or @xmath46 .
so we will simply write @xmath49 instead of @xmath45 unless it is necessary to specify @xmath44 ( _ e.g. _ in the statement of a general theorem like 1.2 ) .
next we consider the case when @xmath11 is non - orientable and the curve @xmath29 is orientation preserving and non - revesible . in this case , we need to fix a lift @xmath398\to\widetilde m$ ] of the chosen closed curve @xmath399\to m$ ] .
assume that the curve @xmath29 is freely homotopic to @xmath44 as closed curves , and let @xmath400 be a free homotopy which keeps the curve closed .
then @xmath388 lifts to a free homotopy @xmath401 from @xmath46 to some lift @xmath21 of @xmath29 .
let @xmath14 be the initial point @xmath308 of @xmath21 .
free winding number _
@xmath45 of the regular closed curve @xmath29 above is defined to be @xmath31 .
@xmath45 is well - defined .
suppose there are two free homotopies @xmath400 and @xmath402 . these lift to homotopies @xmath403 and @xmath404 .
let @xmath358 denote the initial point @xmath365 of @xmath342 .
we want to show @xmath405 .
the composite of @xmath388 and @xmath253 determines a closed curve @xmath280 based at the base point @xmath15 of @xmath29 , and the composed homotopy can be regarded as a homotopy between the ` conjugate ' @xmath386 and @xmath29 which fixes the base point . since @xmath29 is non - reversible , @xmath280 must be orientation preserving .
therefore @xmath406 . by , 4.13 , we have @xmath405 .
@xmath45 is a regular homotopy invariant of @xmath29 .
we only need to consider the case when @xmath11 is non - orientable and @xmath29 is orientation preserving and non - reversible .
suppose @xmath29 and @xmath390 are regularly homotopic regular closed curves . as in the definition above , let @xmath21 be the lift of @xmath29 determined by the lift of a homotopy @xmath407 which starts from @xmath46 .
then a free homotopy between @xmath44 and @xmath390 is obtained by composing @xmath388 and a regular homotopy @xmath253 between @xmath29 and @xmath390 .
now the desired lift @xmath392 which is used to define @xmath408 is determined by the lift @xmath409 starting from @xmath21 . by 5.1 , @xmath410 , where @xmath14 and @xmath411 denotes the initial points of @xmath21 and @xmath392 , respectively .
this completes the proof .
* proof of theorem 1.2 .
* we only need to prove that ( 2 ) implies ( 1 ) .
so assume that @xmath39 and @xmath40 are both freely homotopic to @xmath44 keeping the curve closed and that their free winding numbers @xmath412 and @xmath413 are the same .
let @xmath376 be the base point of @xmath44 .
we can use finger moves and local rotations so that @xmath39 and @xmath40 have the same base point @xmath376 and the same base direction .
these operations are regular homotopy , so the free winding numbers are unchanged . since they are freely homotopic , there exists a smooth closed curve @xmath280 based at @xmath15 such that if we regularly homotope @xmath40 by a finger move along @xmath280 then @xmath39 and the new @xmath40 are homotopic fixing the base point
. let @xmath14 be a lift of @xmath15 in the universal cover @xmath12 , and @xmath303 and @xmath304 be the lifts of @xmath39 and @xmath40 with the initial point @xmath14 , respectively .
since @xmath132 s are homotopic fixing the base point , @xmath307 s have the same terminal point . note that , if @xmath11 is non - orientable and @xmath132 s are reversible in the sense of 5.4 , then we have @xmath414 and that , otherwise , we have @xmath43 .
so suppose @xmath11 is non - orientable , @xmath132 s are reversible , and @xmath415 .
then a finger move of @xmath40 along a suitably chosen closed loop reverses the sign of @xmath416 and we have @xmath43 .
now , by 1.1 , @xmath39 and @xmath40 are regularly homotopic fixing the base point .
this finishes the proof of theorem 1.2 .
@xmath357 since any shortest closed geodesic in a non - trivial free homotopy class of closed curves is a regular closed curve , _
i.e. _ its external angle at the base is 0 ; therefore , its free winding number is 0 .
so we have the following : let @xmath29 be a regular closed curve with the free winding number @xmath326 or @xmath417 . then @xmath29 is regularly homotopic to the following curve : * [ null - homotopic case ] * * ( @xmath418 case ) a figure eight curve , * * ( @xmath419 case ) a small circular curve which winds around a point @xmath215 times , * [ homotopically non - trivial case ] * * ( @xmath418 case ) a closed geodesic in the same free homotopy class , * * ( @xmath419 case ) a curve obtained from a closed geodesic in the same free homotopy class by adding @xmath215 small twists .
we close this section by checking the dependence of the free winding number on the choice of @xmath46 .
assume that @xmath11 is non - orientable and that @xmath44 is an orientation preserving and non - reversible closed curve with base point @xmath376 .
\(1 ) suppose that @xmath420 and @xmath421 are lifts of @xmath376 in @xmath12 , and @xmath46 and @xmath422 are the lifts of @xmath44 with the initial point @xmath423 and @xmath421 , respectively .
then we have @xmath424 \(2 ) suppose that there is a free homotopy @xmath253 from another closed curve @xmath39 to @xmath44 which keeps the curve closed and suppose that a lift @xmath409 of @xmath253 is a homotopy from @xmath303 to @xmath46
. then we have @xmath425 let @xmath388 , @xmath426 , @xmath21 and @xmath14 be as in 5.6 .
\(1 ) let @xmath106 denote the deck transformation which sends @xmath423 to @xmath421 .
then @xmath427 is a homotopy between @xmath428 and @xmath429 ; therefore , we have @xmath430 \(2 ) if we compose the two homotopies @xmath409 and @xmath426 , then it is a homotopy from @xmath303 to @xmath21 and it covers the composition of @xmath253 and @xmath388 . therefore , @xmath45 and @xmath431 are both equal to @xmath31 .
theorem 1.2 contains the classical whitney s theorem for regular closed curves on the euclidean plane . in this section
we consider the other euclidean cases : @xmath432 , @xmath274 , @xmath275 , @xmath276 with the standard tame isogonally euclidean structures .
more precisely , we describe the open annulus @xmath272 and the open mbius band as quotients of the half open unit square @xmath433\times(-\frac{1}{2},\frac{1}{2})$ ] and the torus @xmath275 and the klein bottle @xmath276 as quotients of the unit square @xmath433\times[-\frac{1}{2},\frac{1}{2}]$ ] whose edges are identified as in the pictures below .
the dotted lines show the @xmath5-axis .
the next pictures show their universal covers .
dots show the preimage of a typical point marked in the picture above , and the dotted lines show the @xmath5-axis . to describe the free homotopy class of a curve @xmath3\to m$ ] on @xmath11 , we introduce two indices .
let @xmath15 be the base point of @xmath29 and @xmath14 be a lift of @xmath15 in @xmath12 .
choose a lift @xmath16\to\widetilde m$ ] so that @xmath17 , and let @xmath434 be the expression using the @xmath5- and @xmath120-coordinates of @xmath2 .
\(1 ) the _ horizontal index _
@xmath435 of @xmath29 is defined to be @xmath436 .
this is independent of the choice of @xmath14 , and it completely determines the free homotopy class of @xmath29 when @xmath432 or @xmath274 . for @xmath437 and @xmath276 ,
the curve @xmath29 is orientation preserving if and only if @xmath438 is even .
\(2 ) when @xmath439 , we define the _ vertical index
_ @xmath440 of @xmath29 to be @xmath441 .
this is independent of the choice of @xmath14 , and this together with @xmath438 completely determines the free homotopy class of @xmath29 .
\(3 ) when @xmath442 , we define the _ vertical index
_ @xmath443 of @xmath29 as follows : * if @xmath438 is even ( _ i.e. _ if @xmath29 is orientation preserving ) , we define @xmath444 to be @xmath445 , and * if @xmath438 is odd ( _ i.e. _ if @xmath29 is orientation reversing ) , we define @xmath446 to be @xmath447 modulo 2 .
@xmath443 is obviously invariant under free homotopy .
two closed curves @xmath448\to kb$ ] on @xmath276 are freely homotopic if and only if they have the same horizontal index and the same vertical index .
suppose @xmath449 and @xmath450 .
we may assume that they have the same base point @xmath451\in kb$ ] .
let @xmath14 be a lift of @xmath15 , and take lifts @xmath452\to\widetilde{kb}$ ] of @xmath39 and @xmath40 such that @xmath453 .
let @xmath454 and @xmath455 . by the assumption , we have an equality @xmath456 .
first suppose that @xmath132 s are orientation preserving .
then , we have : @xmath457 if @xmath458 holds , then @xmath459 ; therefore , @xmath460 and @xmath461 have the same end points .
since @xmath462 is contractible , there is a homotopy fixing the end points between @xmath460 and @xmath461 , and this homotopy induces a homotopy between @xmath39 and @xmath40 .
if @xmath463 , then we have @xmath464 .
move @xmath40 on @xmath274 horizontally to the right so that the covering homotopy shifts @xmath304 horizontally by one unit .
on @xmath276 the base point of @xmath40 comes back to the same point , but the curve itself becomes upside down .
therefore , if @xmath455 denotes the lift of the new @xmath40 whose initial point is @xmath14 , then we have @xmath458 ; thus we can proceed as before to obtain a homotopy between @xmath39 and @xmath40 .
next suppose that the curves are orientation reversing .
then we have @xmath465 since @xmath466 , we have @xmath467 .
now move the base point of @xmath40 vertically by one unit either upward or downward , then one of the end points of @xmath304 moves up by one unit and the other end point moves down by one unit ; _ i.e. _ the difference @xmath468 increases by 2 or decreases by 2 .
continue this operation choosing appropriate moves until we have the identity @xmath469 .
now replace the lift @xmath304 by the lift which starts at @xmath14 .
the new lift @xmath304 is parallel to the old lift ; now , @xmath303 and @xmath304 have the same end points , and we can construct a desired homotopy .
we give an explicit formula for @xmath31 in terms of the @xmath20-indices or @xmath25-indices of a lift @xmath21 of @xmath29 .
let @xmath11 be @xmath272 , @xmath274 , @xmath275 , or @xmath276 with the standard tame isogonally euclidean structure , @xmath3\to m$ ] be a regular closed curve based at @xmath15 , and @xmath470 be a lift of @xmath29 whose initial point is a lift @xmath14 of @xmath15 .
when @xmath30\ne 1\in\pi_1(m , p)$ ] , @xmath33 denotes the shortest geodesic on @xmath11 based at @xmath15 that represents the class @xmath30 $ ] .
\(1 ) if @xmath29 is orientation preserving , then @xmath311 .
\(2 ) if @xmath29 is orientation reversing , then @xmath471 .
\(3 ) if @xmath29 is orientation preserving and @xmath30\ne 1 $ ] , then @xmath472 .
\(4 ) if @xmath29 is orientation reversing , then @xmath473 \(5 ) ( formula for @xmath31 ) @xmath474
\(1 ) this holds because @xmath475 .
\(2 ) this holds because @xmath476 .
\(3 ) in this case , the geodesic @xmath33 is a regular closed curve , and we have @xmath477 and @xmath478 .
\(4 ) let @xmath479 be the lift of @xmath33 satisfying @xmath480 , and let @xmath481 .
the pictures below show the situations in the case @xmath482
. we can conclude that @xmath483 , and hence @xmath484 .
the pictures below show the situations in the case @xmath485 .
we can conclude that @xmath486 , and hence @xmath487 .
\(5 ) this follows from ( 1)(4 ) .
next we give an explicit formula for the free winding number @xmath45 in terms of the @xmath20-indices or @xmath25-indices of a lift @xmath21 of @xmath29 .
\(1 ) let @xmath11 be @xmath274 or @xmath276 .
if @xmath3\to m$ ] is an orientation preserving regular closed curve on @xmath11 , and @xmath488 , then @xmath29 is reversible in the sense of 5.3 if and only if @xmath489 .
\(2 ) any orientation preserving closed curve of @xmath274 is reversible .
\(3 ) an orientation preserving non - reversible curve @xmath29 on @xmath276 have two types of lifts in @xmath490 : those satisfying @xmath491 ( called ` positive ' lifts ) and those satisfying @xmath492 ( called ` negative ' lifts ) .
any two lifts of the same type are parallel translations of each other , and their initial points have the same orientation . on the other hand ,
two lifts of different types are parallel translations of a reflection of each other and their initial points have the opposite orientations .
\(4 ) if @xmath388 is a free homotopy between two orientation preserving non - reversible curves @xmath44 and @xmath29 on @xmath276 which keeps the curve closed , and if @xmath388 lifts to a homotopy between @xmath46 and @xmath21 in @xmath490 , then @xmath46 and @xmath21 are parallel translations of each other .
therefore , @xmath46 is a positive lift if and only if @xmath21 is a positive lift .
so , @xmath45 can be calculated using any positive ( _ resp . _
negative ) lift @xmath21 of @xmath29 if @xmath46 is a positive ( _ resp . _ negative ) lift of @xmath44 .
suppose @xmath44 is an orientation preserving non - reversible closed curve on @xmath274 .
if @xmath46 is a positive lift of @xmath44 , then @xmath1 is denoted @xmath493 or simply @xmath49 .
if @xmath46 is a negative lift of @xmath44 , then @xmath1 is denoted @xmath494 .
of course , @xmath495 .
let @xmath11 be @xmath272 , @xmath274 , @xmath275 or @xmath276 with the standard tame isogonally euclidean structure , @xmath3\to m$ ] be a regular closed curve based at @xmath15 , and @xmath488 be a lift of @xmath496 .
\(1 ) if @xmath11 is orientable , @xmath497 ( 2 ) if @xmath11 is non - orientable and @xmath29 is orientation preserving , @xmath498 ( 3 ) if @xmath11 is non - orientable and @xmath29 is orientation reversing , @xmath499 consider the curve on @xmath276 whose lifts in @xmath490 are illustrated below .
there are two types of lifts : those colored in black and those colored in red .
let us use the thick black lift .
obviously @xmath500 .
@xmath501 . set the initial angle to be @xmath502 ; then , the terminal angle is @xmath503 ; therefore , @xmath504 .
we give a combinatorial formula for the @xmath20-indices of regular plane curves with distinct end points .
assume that a regular curve @xmath505\to{\mathbb r}^2 $ ] is given the arc length parameter @xmath65 , where @xmath66 is the length of @xmath29 .
the velocity vector @xmath506 has length 1 .
define a new regular curve @xmath507 by @xmath508 and , for a positive number @xmath509 , define @xmath510\to{\mathbb r}$ ] to be the restriction of @xmath511 to @xmath512 $ ] .
after a reparametrization , @xmath513 is regularly homotopic to @xmath29 , moving the end points but fixing the end directions . in particular , the @xmath20-indices of @xmath513 and @xmath29
are the same . * if @xmath519 , then @xmath516 is the unique angle from @xmath517 to @xmath518 such that @xmath520 , * otherwise , @xmath516 is @xmath521 or @xmath522 according as the point @xmath523 sits on the left - hand side or the right - hand side of the line through @xmath524 with direction @xmath517 .
first , let us consider the @xmath519 case .
if @xmath509 is very large and if we look at the curve @xmath511 from far away , then the original curve @xmath29 is too small to recognize , and the curve looks like this : note that the whole curve lies on the left - hand side of the tangent line at @xmath15 ; therefore , we can apply whitney s
formula @xcite to express the @xmath20-index @xmath532 of the curve @xmath531 using the information of the double points as follows : @xmath534 the desired formula immediately follows from this . | we introduce winding numbers @xmath0 and free winding numbers @xmath1 of regular closed curves on surfaces with a nice euclidean or hyperbolic geometry .
such surfaces include the euclidean plane , the annulus , the mbius band , the torus , the klein bottle , closed hyperbolic surfaces , and complete hyperbolic surfaces .
we show that ( 1 ) regular closed curves with the same base point are regularly homotopic fixing the base point if and only if they represent the same element of the fundamental group and have the same winding number @xmath0 and ( 2 ) two regular closed curves are regularly homotopic if and only if they are freely homotopic and have the same free winding number @xmath1 .
these two winding numbers are integers if the curve is orientation preserving , and are integers modulo 2 if the curve is orientation reversing .
we also give a combinatorial formula for the ` winding numbers ' of regular nonclosed plane curves , which is useful for computing these winding numbers of closed curves . |
the diffuse interstellar medium ( ism ) contains gas over a wide range of densities , temperatures and ionization states
. these can be broadly sub - divided into the molecular phase ( e.g. @xcite ) , the neutral atomic phase ( e.g. @xcite ) , and the warm and hot ionized phases ( e.g. @xcite ) . the neutral atomic medium ( mostly neutral hydrogen , hi ) is further usually sub - divided into `` cold '' and `` warm '' phases ( the `` cnm '' and `` wnm '' , respectively ) .
typical cnm temperatures and densities are @xmath6 k and @xmath7 @xmath8 , respectively , with corresponding wnm values of @xmath9 k and @xmath10 @xmath8 .
this was originally an observational definition , to distinguish between phases producing strong narrow absorption lines towards background radio - loud quasars and smooth broad emission lines @xcite .
later , this separation into cold and warm phases was found to arise naturally in the context of models in which the atomic and ionized phases are in pressure equilibrium .
atomic gas at intermediate temperatures was found to be unstable to perturbations , and to migrate to either the cold or warm stable phases , given sufficient time to attain pressure equilibrium @xcite .
radio studies of the hi-21 cm transition have played a vital role in understanding physical conditions in the neutral ism .
the hi-21 cm excitation temperature of a gas cloud ( the `` spin temperature '' , @xmath11 ) can be shown to depend on the local kinetic temperature ( @xmath12 ; @xcite ) , with @xmath13 in the cnm and @xmath14 in the wnm @xcite .
since the hi-21 cm absorption opacity for a fixed hi column density varies inversely with @xmath11 , while the emissivity is independent of @xmath11 ( in the low opacity limit ) , hi-21 cm absorption studies against background radio sources are primarily sensitive to the presence of cnm along the sightline , while hi-21 cm emission studies are sensitive to both warm and cold hi .
indeed , the original evidence for two temperature phases in the neutral gas stemmed from a comparison between hi-21 cm absorption and emission spectra @xcite .
physical conditions in the neutral gas are also known to depend on the gas column density . at the low hi column densities of the inter - galactic medium , @xmath15 @xmath16 ,
the gas is optically thin to ionizing ultraviolet ( uv ) photons and is hence mostly ionized , giving rise to the so - called lyman-@xmath5 forest @xcite . at higher hi column densities , @xmath17 @xmath16 ,
the hi is optically - thick to ionizing photons at the lyman - limit , and the core of the lyman-@xmath5 absorption line is saturated .
such `` lyman - limit systems '' are partially ionized and typically arise for sightlines through the outskirts of galaxies @xcite . at still higher hi column densities , @xmath18 @xmath16 ,
the lyman-@xmath5 line becomes optically - thick in its naturally - broadened wings and acquires a lorentzian shape ; such systems are called damped lyman-@xmath5 absorbers ( dlas ; @xcite ) .
damped absorption is ubiquitous on sightlines through galaxy disks and , at high redshifts , has long been used as the signature of the presence of an intervening galaxy along a quasar sightline @xcite .
finally , the molecular hydrogen fraction shows a steep transition from very low values to greater than 5% at @xmath19 @xmath16 in the galaxy @xcite ) .
neutral gas is likely to become predominantly molecular at higher hi column densities , @xmath20 @xmath16 @xcite . while it is well known that a threshold column density ( @xmath19 @xmath16 ) is required to form the molecular phase , for self - shielding against uv photons in the @xmath21 lyman band @xcite , the conditions for cnm formation are less clear . in this _ letter _ , we report results from hi-21 cm absorption studies of a large sample of compact radio sources that indicate the presence of a similar , albeit lower , hi column density threshold for cnm formation in the milky way .
we have used the westerbork synthesis radio telescope ( wsrt , 23 sources ) , the giant metrewave radio telescope ( gmrt , 10 sources ) and the australia telescope compact array ( atca , 2 sources ) to carry out sensitive , high spectral resolution ( @xmath22 km / s ) hi-21 cm absorption spectroscopy towards 35 compact radio - loud quasars .
most of the target sources were selected to be b - array calibrators for the very large array , and have angular sizes @xmath23 .
details of the sample selection , observations and data analysis are given in @xcite , @xcite , roy et al .
( 2011 , in prep . ) and kanekar & braun ( 2011 , in prep . ) .
the final optical - depth spectra have root - mean - square ( rms ) noise values of @xmath24 per 1 km / s channel , with a median rms noise of @xmath25 per 1 km / s channel .
these rms noise values are at off - line channels , away from hi-21 cm emission that increases the system temperature , and hence , the rms noise . a careful bandpass calibration procedure was used to ensure a high spectral dynamic range and excellent sensitivity to wide absorption [ see @xcite and @xcite for details ] .
these are among the deepest hi-21 cm absorption spectra ever obtained ( e.g. @xcite ) and constitute by far the largest sample of absorption spectra of this sensitivity .
the use of interferometry also implies that the spectra are not contaminated by hi-21 cm emission within the beam , unlike the case with single - dish absorption spectra @xcite .
galactic hi-21 cm absorption was detected against every source but one , b0438 - 436 .
we used the leiden - argentine - bonn ( lab ) survey @xcite to estimate the `` apparent '' hi column density ( uncorrected for self - absorption , i.e. @xmath26 @xmath16 @xcite , where @xmath27 is the brightness temperature ) at a location adjacent to each target .
the observational results are summarized in table [ tab : summary ] , whose columns contain ( 1 ) the quasar name , ( 2 ) its galactic latitude , ( 3 ) the hi column density ( uncorrected for self - absorption ) , in units of @xmath28 @xmath16 , from the lab survey , ( 4 ) the integrated hi-21 cm optical depth @xmath29 , in km / s , and ( 5 ) the harmonic - mean spin temperature @xmath11 in k , estimated assuming the low optical depth limit ( @xmath30 ) .
finally , for the one source with undetected hi-21 cm absorption ( b0438 - 436 ) , we quote a @xmath31 upper limit on the integrated hi-21 cm optical depth , assuming a gaussian line profile with a full width at half maximum of 20 km / s , and the corresponding @xmath31 lower limit on the spin temperature .
.the full sample of 35 sources .
[ tab : summary ] [ cols="^,^,^,^,^ " , ] 0.01 in notes : @xmath32 from the lab emission survey .
@xmath33 from our hi-21 cm absorption survey .
the left panel of figure [ fig : fig1 ] shows the integrated hi-21 cm optical depth plotted against hi column density , on a logarithmic scale .
it is clear from the figure that the integrated hi-21 cm optical depth has a rough power - law dependence on @xmath34 for @xmath18 @xmath16 .
however , there is a steep decline in @xmath35 for hi column densities lower than this value .
these properties were noted previously by @xcite for individual spectral features ( see their figs .
36 ) rather than the line - of - sight integral .
a similar pattern is visible in the right panel of the figure , which plots the spin temperature @xmath11 against @xmath34 .
24 out of 25 sightlines with @xmath36 @xmath16 show @xmath37 k , while all ten sightlines with @xmath38 @xmath16 have @xmath39 k. the median spin temperature for sightlines with @xmath18 @xmath16 is @xmath40 k , while that for sightlines with @xmath41 k is @xmath42 k. thus , there appears to be a physical difference between sightlines with hi columns lower and higher than the column density @xmath43 @xmath16 .
note that the use of apparent @xmath34 only affects high - opacity sightlines , and has no significant effect on our results .
we used a number of non - parametric two - sample tests ( e.g. the gehan test , the peto - prentice test , the logrank test , etc ) , generalized for censored data , to determine whether the spin temperatures of sightlines with @xmath44 and @xmath45 are drawn from the same parent distribution .
all tests used the asurv rev .
1.2 package @xcite which implements the methods of @xcite .
the use of multiple tests guards against any biases within a given test , resulting from the relatively small sample size ( 10 systems with @xmath45 ) and the presence of censored values @xcite .
errors on individual measurements were handled through a monte - carlo approach , using the measured values of @xmath11 and @xmath34 ( and the associated errors ) for each sightline to generate @xmath46 sets of 35 pairs of @xmath11 and @xmath34 values .
the statistical significance of each result ( quoted below ) is the average of values obtained in these @xmath46 trials for each test .
the tests found clear evidence , with statistical significance between @xmath47 and @xmath48 that the spin temperatures for sightlines with @xmath44 and @xmath45 are drawn from different parent populations .
following @xcite , our final result is based on the peto - prentice test , as this has been found to give the best results for very different sample sizes @xcite .
the two @xmath11 sub - samples ( with `` low '' and `` high '' @xmath34 ) are then found to be drawn from different parent distributions at @xmath49 significance ; the probability of this arising by chance is @xmath50 .
it thus appears that sightlines with @xmath51 @xmath16 have systematically higher spin temperatures than sightlines with @xmath52 .
the spin temperature measured here is the column - density - weighted harmonic mean of the spin temperatures of different `` phases '' along the line of sight ( e.g. @xcite ) .
for example , a sightline with @xmath34 equally divided between cnm and wnm , with @xmath53 k and @xmath54 k respectively , would yield an average @xmath55 k , while one with 90% of gas with @xmath56 k and 10% with @xmath53 k would yield an average @xmath57 k. in other words , a high @xmath11 indicates the presence of a smaller fraction of the cold phase of hi .
thus , the fact that average spin temperatures are significantly higher on sightlines with low hi column densities , @xmath38 @xmath16 , indicates that such low-@xmath34 sightlines contain low cnm fractions , far smaller than those on sightlines with @xmath18 @xmath16 .
it is clear from fig .
[ fig : fig1][a ] that the integrated hi-21 cm optical depth drops sharply at @xmath58 , due to the decline in the cnm fraction at low hi column densities .
this can be accounted for by a simple physical model in which a minimum `` shielding '' hi column density of wnm is needed for the formation of the cold phase in an hi cloud .
assuming that the hi-21 cm optical depth in the wnm is negligible compared to that in the cnm , the equation of radiative transfer for a `` sandwich '' geometry can be written as @xmath59 where the explicit velocity dependence of @xmath60 and @xmath61 has been omitted for clarity and the integrated hi emission profile becomes , @xmath62 or , @xmath63 for a measured `` apparent '' column density , @xmath64 ( assuming negligible self - opacity ) , a threshold column density ( where @xmath65 ) , @xmath66 , a saturation column density ( where @xmath67 ) , @xmath68 and an effective opacity , @xmath69 . the effective opacity is related to the measured integrated opacity by the effective linewidth , @xmath70 , as @xmath71 . the upper and lower dashed curves in fig .
[ fig : fig1][a ] show the above expression for ( @xmath72 @xmath16 , @xmath73 @xmath16 , @xmath74 km / s ) and ( @xmath75 @xmath16 , @xmath76 @xmath16 , @xmath77 km / s ) , respectively . note that the effective linewidth only shifts the curves up and down in the figure .
the fact that none of the data points of fig .
[ fig : fig1][a ] lie to the left of the upper curve indicates that a minimum _ shielding _ hi column of @xmath78 \sim 5 \times 10^{19}$ ] @xmath16 is needed for the formation of the cold neutral medium . note that , in the above `` sandwich '' geometry , a total wnm column density of @xmath79 @xmath16 is needed before any cnm can be formed
most theoretical models of the diffuse interstellar medium treat it as a multi - phase medium , consisting of neutral and ionized phases in pressure equilibrium ( e.g. @xcite ) . in the mckee - ostriker model , physical conditions in the ism are regulated by supernova explosions , whose blast waves sweep up gas in the ism into shells , leaving large cavities full of hot ionized gas @xcite .
cold atomic gas is then formed by the rapid cooling of the swept - up and shocked gas .
soft x - rays from neighbouring hot ionized gas and stellar uv photons penetrate into the cnm and heat and partially ionize it , producing envelopes of warm neutral and warm ionized gas ( wim ) .
thus , a typical ism `` cloud '' [ see fig . 1 of @xcite ] is expected to form at the edges of supernova remnants and super - bubbles containing the hot ionized medium ( him ) .
such a cloud consists of a cnm core surrounded by a wnm envelope and a further wim shell , all of which are in pressure equilibrium .
the cnm core is almost entirely neutral , the wnm and wim are partially ionized , and the him is almost entirely ionized @xcite .
a threshold hi column density for cnm formation in an ism cloud arises quite naturally due to the need for self - shielding against ionizing ultraviolet ( and soft x - ray ) photons ( see also @xcite ) . at low hi columns
, uv photons can penetrate all the way into a cloud core and heat ( and partly ionize ) the hi . only clouds
that self - shield against the penetration of these high - energy photons can retain a stable cnm core .
self - shielding against uv photons is significant once the optical depth at the lyman limit crosses unity , i.e. for @xmath80 @xmath16 , and becomes more and more efficient with increasing hi column density .
our results indicate that self - shielding only becomes entirely efficient at excluding uv photons from the interior of hi clouds at a _ total _ hi column density of @xmath81 @xmath16 . below this threshold ,
sufficient uv photons penetrate into the cloud interior to hinder the survival of the cold phase .
the simple `` sandwich '' model overlaid on fig .
[ fig : fig1][a ] suggests that a shielding column of @xmath82 @xmath16 _ on each exposed surface of a cloud _ is sufficient to exclude uv photons , with the rest of the hi then free to cool to lower temperatures .
it is also possible that sightlines with low @xmath34 are sampling gas at higher distances from the galactic plane , with lower metallicity and pressure .
two - phase models that incorporate detailed balancing of heating and cooling rates obtain lower cnm fractions at lower pressures and metallicities @xcite .
observationally , @xcite have found evidence for an anti - correlation between spin temperature and metallicity in high-@xmath83 dlas , with low - metallicity dlas having higher spin temperatures , probably because the paucity of metals yields fewer cooling routes @xcite . if low-@xmath34 sightlines contain `` clouds '' with systematically lower metallicities ( and/or pressures )
, these could have lower cooling rates and hence , lower cnm fractions .
unfortunately , we do not have estimates of either pressure or metallicity along our sightlines and hence can not test this possibility .
we note that all sightlines with high spin temperatures are at intermediate or high galactic latitudes , @xmath84 ( see fig .
[ fig : fig2 ] ) , suggesting larger distances from the plane .
conversely , the figure also shows that low spin temperatures ( @xmath85 k ) are obtained even for sightlines at high galactic latitudes ( @xmath86 ) .
measurements of the metallicity and pressure along the low-@xmath34 sightlines , through uv spectroscopy , would be of much interest .
there have been earlier suggestions , based on semi - analytical or numerical treatments of self - shielding , that hi clouds are predominantly neutral for @xmath87 @xmath16 ( e.g. @xcite ) . however , this is the first direct evidence for a change in physical conditions in hi clouds at this column density .
note that this is a factor of several lower than the known threshold of @xmath19 @xmath16 for the formation of molecular hydrogen in galactic clouds , set by the requirement of self - shielding against uv photons at wavelengths in the h@xmath88 lyman band @xcite .
thus , there appear to be three column densities at which phase transitions occur in ism clouds , at @xmath81 @xmath16 resulting in the formation of cold hi , at @xmath89 @xmath16 resulting in the formation of molecular hydrogen , and finally , at @xmath90 @xmath16 , when most of the atomic gas is converted into the molecular phase . in this context , it is vital to appreciate that the abscissa of fig .
[ fig : fig1 ] refers to _ apparent _ hi column density under the assumption of negligible self - opacity of the hi profile .
the saturation column density , n@xmath91 , is an _
observational _ upper limit to @xmath92 and does not represent a physical limit on @xmath34 .
in fact , the widespread occurrence of the hi self - absorption phenomenon within the galaxy @xcite and the detailed modeling of high resolution extragalactic hi spectra suggests that self - opacity is a common occurrence @xcite which can disguise neutral column densities that reach @xmath93 @xmath16 .
the original definition of a dla as an absorber with @xmath94 @xmath16 was an observational one , based on the requirement that the damping wings of the lyman-@xmath5 line be detectable in low - resolution optical spectra of moderate sensitivity @xcite . with today
s 10-m class optical telescopes , it is easy to detect the damping wings at significantly lower hi column densities , @xmath95 @xmath16(e.g .
@xcite ) ; such systems are referred to as `` sub - dlas '' .
there has been much debate in the literature on whether or not dlas and sub - dlas should be treated as a single class of absorber and on their relative importance in contributing to the cosmic budget of neutral hydrogen and metals ( e.g. @xcite ) .
@xcite argue that dlas and sub - dlas are physically different , claiming that most of the hi in sub - dlas is ionized and at high temperature , while that in dlas is mostly neutral ; this is based on numerical estimates of self - shielding in dlas and sub - dlas against the high-@xmath83 uv background @xcite .
our detection of an hi column density threshold for cnm formation that matches the defining dla column density indicates that dlas and sub - dlas are indeed physically different types of absorbers , with sub - dlas likely to have significantly lower cnm fractions than dlas , at a given metallicity . in summary ,
we report the discovery of a threshold hi column density , @xmath96 @xmath16 , for cold gas formation in hi clouds in the interstellar medium . above this threshold ,
the majority of galactic sightlines have low spin temperatures , @xmath97 k , with a median value of @xmath40 k. below this threshold , typical sightlines have far higher spin temperatures , @xmath98 k , with a median value of @xmath42 k. the threshold for cnm formation appears to arise naturally due to the need for self - shielding against ultraviolet photons , which penetrate into the cloud at lower hi columns and heat and ionize the hi , inhibiting the formation of the cold neutral medium .
we thank the staff of the gmrt and wsrt for help during the observations .
the gmrt is run by the national centre for radio astrophysics of the tata institute of fundamental research .
the wsrt is operated by astron with support from the netherlands foundation for scientific research ( nwo ) .
nk acknowledges support from the department of science and technology through a ramanujan fellowship .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . | we report the discovery of a threshold in the hi column density of galactic gas clouds below which the formation of the cold phase of hi is inhibited .
this threshold is at @xmath0 per @xmath1 ; sightlines with lower hi column densities have high spin temperatures ( median @xmath2 k ) , indicating low fractions of the cold neutral medium ( cnm ) , while sightlines with @xmath3 per @xmath1 have low spin temperatures ( median @xmath4 k ) , implying high cnm fractions .
the threshold for cnm formation is likely to arise due to inefficient self - shielding against ultraviolet photons at lower hi column densities .
the threshold is similar to the defining column density of a damped lyman-@xmath5 absorber ; this indicates a physical difference between damped and sub - damped lyman-@xmath5 systems , with the latter class of absorbers containing predominantly warm gas . |
the study of topological phases of matter has considerably grown in recent years and new systems have been investigated both theoretically and experimentally @xcite .
it is now apparent that some remarkable topological features of quantum hall states can occur in a wider set of systems and thus be more universal and robust . in particular , non - chiral topological states , such as those of the quantum spin hall effect , of topological insulators and of topological superconductors , do not require strong magnetic fields and exist in three space dimensions @xcite@xcite . a characteristic feature of topological states is the existence of massless edge excitations that are well accounted for by low - energy effective field theory descriptions @xcite .
the response of topological states to external disturbances does not occur in the gapped bulk , but manifests itself through the edge dynamics .
while the edge excitations of chiral states , such as quantum hall states and chern insulators , are absolutely stable , those of non - chiral topological states can interact and became gapful , leading to the decay into topologically trivial phases . in some cases ,
edge interactions are forbidden by the presence of ( discrete ) symmetries : we then speak of symmetry protected topological phases of matter @xcite .
topological insulators protected by time - reversal ( tr ) symmetry have been first analyzed in free fermion systems using band theory @xcite@xcite .
these systems were found to be characterized by a topological bulk quantity equal to the @xmath0 index @xmath1 , where @xmath2 is the number of fermion edge modes of each chirality .
the odd ( even ) @xmath2 cases were shown to be stable ( unstable ) upon using a remarkable flux argument by fu , kane and mele @xcite .
the stability analysis was then reformulated in terms of the dynamics of edge excitations in refs .
@xcite@xcite@xcite , by using conformal field theory ( cft ) methods @xcite .
it was shown that the @xmath0 classification extends to interacting systems that are described by abelian edge excitations within the so - called @xmath3-matrix formalism @xcite .
the analysis of levin and stern led to the following simple and general result for the @xmath0 index @xcite : ( -1)^2s , 2s= , [ ls - index ] where @xmath4 is the spin hall conductance and @xmath5 is the minimal fractional charge for one of the spin ( chiral ) components , in units of @xmath6 and @xmath7 , respectively .
their ratio measures the spin @xmath8 of an excitation created at the edge , as explained later .
an odd ( even ) ratio corresponds to stable ( unstable ) systems , as this quantity reduces to the number of fermion modes in the non - interacting case . in this paper , we rederive and further extend the @xmath0 stability criterion ( [ ls - index ] ) by studying the partition function of edge excitations . in earlier work , we obtained the general form of the partition function for both abelian and non - abelian quantum hall states @xcite@xcite@xcite . upon generalizing
this analysis to the quantum spin hall effect and topological insulators , we can apply the flux argument for stability to any type of interacting topological insulator and prove the general validity of ( [ ls - index ] ) .
our analysis clarifies that the stability is associated to the presence of an anomaly in the @xmath9 symmetry of fermion number parity at the edge , also equal to the edge spin parity @xmath10 .
actually , the @xmath11 spin symmetry of the quantum spin hall effect is explicitly broken to @xmath10 by spin - orbit interaction and other tr invariant relativistic corrections that are present in generic topological insulators @xcite .
the @xmath9 anomaly is thus the remnant of the @xmath11 anomaly of the spin hall effect , which is analogous to the @xmath12 charge anomaly in the hall effect @xcite .
the partition function is obtained in the double periodic geometry of the torus made by one circular edge and compact euclidean time for temperature @xcite@xcite .
this function contains anyonic sectors for each chirality , that involve sums over charged and neutral edge modes .
this structure allows one to disentangle the abelian charged mode that determines the @xmath0 anomaly , from the neutral modes , either abelian or non - abelian , that are transparent .
we can thus analyze the stability ( anomaly ) for general interacting system .
the response of the topological state , i.e. of its edge dynamics , to external perturbations can be easily described by using the partition function .
one can see how its different sectors transforms among themselves by inserting fluxes , i.e. under the effect of an electromagnetic background . furthermore , the behaviour under modular transformations , the discrete coordinate changes respecting the periodicities of the torus , yield the response to ( some type of ) gravitational backgrounds : they amount , e.g. to adding momentum to the system or to swapping space and time ( respectively , the @xmath13 and @xmath14 modular transformations ) .
we find that the stability ( instability ) of non - chiral edge states is associated to the impossibility ( possibility ) of having a modular invariant partition function that is consistent with time reversal symmetry .
the modular non - invariance of stable topological states is actually a kind of gravitational anomaly accompanying the spin parity anomaly . in some systems , we find that flux insertions and modular transformations act in similar way on the four `` spin sectors '' always present in fermionic systems , corresponding to periodic and antiperiodic boundary conditions in space and time @xcite@xcite . in other systems ,
the two backgrounds have different effects owing to the different couplings of charged and neutral modes .
modular non - invariance and stability of topological states have already been discussed in a paper by ryu and zhang @xcite that actually motivated our work .
they considered the topological superconductors , whose neutral majorana edge modes can not be analyzed by flux arguments , and argued that modular non - invariance of the partition function could be used as a criterion of stability .
then they showed that the system of @xmath15 majorana fermions is unstable for @xmath16 modulo @xmath17 , finding agreement with the study of possible interactions @xcite@xcite .
our analysis reproduces this result , but can not presently provide a general approach to interacting topological superconductors .
this open issue is discussed in the conclusions .
the paper is organized as follows . in section two ,
we briefly describe the setting of the problem and recall the flux insertion argument by fu , kane and mele . in section three ,
we introduce the partition function of topological insulators with one edge mode per spin , described by the luttinger liquid cft with central charge @xmath18 .
we discuss the effect of flux insertions and modular transformations , and rederive the @xmath0 anomaly and the stability argument in this context . in section four
, we extend the analysis to general non - chiral edge cfts and discuss some interesting examples . in section five , we present our conclusions .
the appendix contains the details on modular transformations of abelian and non - abelian theories .
we start by recalling the laughlin argument for the quantization of the hall current in the annulus geometry ( see fig . [ fig1](a ) ) @xcite . upon the adiabatic insertion of one quantum of flux @xmath19 , the bulk hamiltonian returns to itself while states in the spectrum drift one into another , leading to the so - called spectral flow .
this amounts to the transport of a charge equal to the hall conductivity between the two edges . & & + _ 0 , h = h , & & qq+q= .
[ nu - hall ] from the point of view of the conformal field theory describing one edge of the annulus , the spectral flow corresponds to the non - conservation of charge , namely a @xmath12 chiral anomaly in two dimensions @xcite@xcite .
the associated index theorem for the integrated anomaly reads : q= _ ^
dt_0 ^ 2r dx _ t j^0_r = f = e n. n. [ int - anomaly ] the last term in this expression is a topological quantity , the first chern class of the two - dimensional electromagnetic field , @xmath20 , whose integer values count the number of inserted flux quanta .
we also recall that the chiral anomaly is an universal effect that is exact for any strength of the interaction @xcite .
we can interpreted the chiral anomaly of the edge as the response of the topological bulk to an electromagnetic background , as well explained in the associated description in terms of the effective hydrodynamic chern - simons theory @xcite .
other chiral topological phases are also associated to anomalies , leading to the breaking of continuous symmetries and to anomalous currents .
these correspond to non - vanishing values of transport coefficients , such as the hall conductivity @xmath21 and/or thermal transport coefficient @xmath22 .
the general analysis of topological phases characterized by electromagnetic , gravitational and mixed chiral anomalies in two and higher dimensions has been carried out in ref .
@xcite ; these results have identified some @xmath23 classes of topological states in @xmath24 dimensions that actually fit within the tenth - fold classification of non - interacting topological systems @xcite , thus extending it to interacting cases .
the next step is to discuss the laughlin argument for the non - chiral topological state of the quantum spin hall effect ( see fig . [ fig1](b ) ) .
consider the system made of two copies of the @xmath25 hall effect having opposite spin and chiralities . the time reversal ( tr ) transformation @xmath26 acts on up and down spin electrons , resp . @xmath27 and @xmath28 , as follows , : _ k _ -k , _ k -_-k , [ tr - action ] thus leaving the system invariant .
the addition of one flux causes the drift of up and down electrons in opposite directions with respect to the fermi surface at each edge ( fig .
[ fig2 ] ) . from the point of view of the cft at one edge ,
say the outer one , the effect is to create a neutral excitation with spin one @xcite : q = q^+ q^=0 , s = - ( - ) = 1 , [ one - flux - qshe ] where @xmath29 , i.e. the spin hall current is equal the hall current of one chiral component ( in appropriate units ) . and @xmath30 edges of the annulus ( dashed - dotted lines).,width=453 ] arguing as in the chiral case , we can characterize this topological state by the spin chiral anomaly @xmath11 corresponding to an integer spin hall conductivity .
however , the quantum spin hall effect is a rather academic model of topological insulator : in general , spin - orbit coupling and other relativistic effects can not be neglected , that only conserve the total angular momentum ; thus , the @xmath11 symmetry is explicitly broken , the spin current is not defined and the spin hall conductivity vanishes . in the following ,
we want to discuss general topological insulators that only possess time reversal symmetry and can not be characterized by anomalies of continuum symmetries ; these represent interacting topological phases of a different kind . in our analysis
we shall mostly keep the spin - quantum hall description of topological insulators , that can be investigated by cft methods , corresponding to the the ideal limit of spin - orbit coupling switched off .
nevertheless , we shall discuss properties that do not rely on the spin being conserved .
a topological insulator with a single free fermion edge mode per spin ( chirality ) is stable because the mass term coupling the two chiralities is forbidden by tr symmetry , as follows : : h_int .
= m ^ _ _ + h.c
. - h_int . .
[ mass - term ] another tr invariant mass term can be written that couples two fermions modes per spin , but a single fermion always remains massless in a system with an odd number of modes @xcite .
of course , if tr symmetry is broken all edge excitations become gapful ( and the insulator trivial ) .
the analysis of gapful instabilities due to more general , non - quadratic interactions compatible with tr can be done in some cases , but we consider here another criterion for stability that is associated with a symmetry and a discrete @xmath0 anomaly .
this is the fu - kane - mele flux insertion argument called the `` spin pump '' ( a cyclic adiabatic process ) @xcite .
we mostly follow the presentation of ref.@xcite . the insertion of magnetic flux breaks tr symmetry , owing to : h^-1 = h .
[ tr - ham ] this relation together with the periodicity @xmath31=h[\phi+\phi_0]$ ] , implies that the bulk hamiltonian is tr invariant for a discrete set of flux values : = 0,,_0 , , .
[ half - flux ] the fu - kane - mele analysis of band insulators let them to define an index called `` tr invariant polarization '' , @xmath32 , that enjoys the following properties : \i ) it is a bulk topological quantity , conserved by tr symmetry .
\ii ) its value is equal to the spin parity ( fermion number ) at the edge , ( -1)^p_= ( -1)^n_+n_=(-1)^2s .
[ s - p ] \iii ) in a stable topological insulator , it changes value between tr invariant points ( [ half - flux ] ) separated by half flux @xmath33 .
the stability argument goes as follows : adding @xmath34 fluxes in the center of the annulus creates a spin one - half excitation at both boundaries , the two cases being related by a tr transformation ( see fig .
[ fig3 ] ) .
each of these excitations has a tr partner locally at the boundary , making a total of four states .
next , one invokes the kramers theorem at the tr - invariant point @xmath35 to argue that the two spin one - half excitations at one edge are degenerate in energy and orthogonal , this degeneracy being robust to any tr symmetric perturbation .
the energy change of the edge ground state as the flux is varied from zero to @xmath36 is shown in fig .
[ fig4 ] . at @xmath36 ,
the evolved ground state @xmath37 necessarily meets with one excited state @xmath38 owing to kramers theorem .
going back to @xmath39 , the excited state must have an energy @xmath40 from the work done by adding a flux quantum in a system of size @xmath30 .
it then follows that the existence of a kramers ( spin one - half ) pair at the edge for @xmath35 implies the presence of a gapless excitation at @xmath39 that is protected by tr symmetry . in the case of two fermion modes , the corresponding spin one excitation created at the boundary
would not be protected by kramers theorem against energy splitting from its tr companions . the argument
then extend to odd and even numbers of fermion modes .
this completes the argument for stability of the topological phase with an odd number of fermion modes , leading to the @xmath0 classification of topological insulators in the free fermion case .
let us add some remarks : \i ) the existence of a kramers pair is signalled by the change of spin parity of the ground state upon adding half flux , = 0 : ( -1)^2s = 1 = : ( -1)^2s=-1 .
[ change s - p ] \ii ) the spin parity is conserved by tr symmetry , being just another way to state the kramers theorem .
this @xmath0 invariance is the remnant of the continuous @xmath11 symmetry of the quantum spin hall effect that gets broken by relativistic effects .
\iii ) at the two tr symmetric points , @xmath41 , the spin parity takes different values without having included tr breaking terms in the hamiltonian .
therefore , this quantity is anomalous . in conclusions , tr invariant topological insulators are associated with the @xmath0 spin parity symmetry @xmath10 that is anomalous .
the full spin symmetry @xmath11 would also be anomalous but it is explicitly broken in general topological insulators .
we note that the fu - kane - mele argument is a generalization of the laughlin argument that makes it apparent the @xmath0 spin parity anomaly .
the goal of this paper is to generalize the stability analysis to interacting systems through the study of partition functions .
we first recall from ref . @xcite the construction of the partition function of the quantum hall effect for the laughlin states , @xmath42 , @xmath43 odd ; next , we generalize it to topological insulators and discuss the @xmath0 anomaly and the stability criterion in this setting .
we consider the grand - canonical partition function of states at the outer edge of the annulus with radius @xmath30 ( see fig . [ fig5 ] ) .
this circle and the time period @xmath44 realize the geometry of the torus .
the energy and momentum of chiral excitations are expressed by the eigenvalues of the virasoro generator @xmath45 , @xmath46 ( we set the fermi velocity @xmath47 for simplicity ) .
the trace over the states with gibbs weight involving chemical and electric potential decomposes into orthogonal sectors @xmath48 , corresponding to a given value of fractional charge plus any number of electrons , @xmath49 , @xmath50 ; there are @xmath43 sectors for @xmath51 .
the partition function for one sector takes the form : k_(,;p)&=&tr_h^( ) & = & _ n ( i2 ( + ) ) ; [ pf ] this function is parameterized by the two complex numbers , = + t , = ( iv_o + ) , [ tau - zeta ] that are the modular parameter @xmath52 and the `` coordinate '' @xmath53 .
@xmath54 is the ratio of the two periods and @xmath55 is the torsion parameter conjugate to momentum @xmath56 ; @xmath53 contains the electric @xmath57 and chemical @xmath58 potentials .
the expression of @xmath59 ( [ pf ] ) involves a sum of characters of the representations of the @xmath60 current algebra of the @xmath18 cft with charges @xmath49 and conformal weight @xmath61 , as is apparent in the exponent of ( [ pf ] ) @xcite .
it can be obtained by canonical quantization @xcite as well as by using cft representation theory supplemented by some physical conditions on the charge and statistics of electron excitations @xcite .
the formula involves the ratio of a theta function with characteristics @xmath62 ( \zeta|p\tau)$ ] and the dedekind function @xmath63 describing particle - hole excitations .
the non - holomorphic prefactor @xmath64 $ ] is explained in ref.@xcite .
altogether , the partition function of one chiral edge is given by the multiplet of functions @xmath59 , for @xmath65 , having periodicity @xmath66 , that correspond to the @xmath43 anyon sectors .
the torus geometry is left invariant by the modular transformations , the discrete coordinate changes consistent with the double periodicity . these act on the modular parameter @xmath52 and the coordinate @xmath53 as follows @xcite : , , a , b , c , d , ad - bc=1 , [ mod.tr ] and span the group @xmath67 .
the modular group is infinite dimensional and is generated by two transformations , @xmath68 and @xmath69 , obeying the relations @xmath70 , where @xmath71 is the charge conjugation matrix , @xmath72 @xcite .
in addition , there are the two periodicities of the coordinate @xmath53 at @xmath52 fixed : @xmath73 and @xmath74 .
the modular transformations belong to the group of two - dimensional diffeomorphisms of the torus , being the global transformations not connected to the identity : they are the `` large '' gauge transformations of the conformal theory placed in a gravitational background . in similar way , the flux insertions are large gauge transformations of the electromagnetic background .
the lack of invariance of the partition function signals the presence of gravitational and gauge anomalies , respectively .
these do not led to inconsistencies if the backgrounds are classical , i.e. not quantized , as in our case , but nevertheless characterize the low - energy physics of the system @xcite .
the multiplet of @xmath59 transforms linearly under the modular group and each generator has physical significance , as we now recall @xcite .
the @xmath14 transformation reads : s : k_ ( , ) = e^i_=1^p s_ k_(,),s_=(i2 ) , [ s_transf ] where @xmath75 is the modular @xmath14-matrix and @xmath76 is an overall phase .
as is well know in the cft literature , the @xmath14 unitary transformation expresses a completeness condition on the spectrum of the theory ; it also determines the fusion rules of excitations via the verlinde formula @xcite .
the invariance under @xmath77 transformation up to a global phase , t^2 : k_(+2 , ) = ( i4h_ ) k_ ( , ) , h_= , [ t2_transf ] implies that the electron excitations being summed up in each anyon sector have odd integer statistics ( half integer conformal dimension ) .
on the other hand , the @xmath78 invariance , u : k_(,+1)= ( i2 / p ) k_ ( , ) , [ u_transf ] implies electrons with integer charge . the @xmath77 and @xmath78 transformations are readily checked from the expression ( [ pf ] ) .
finally , the @xmath79 transformation , v : k_(,+ ) = e^i k_+1 ( , ) , = _ 0 , [ v_transf ] ( with @xmath80 another global phase ) is very important because it realizes the change of electric potential following the addition of one flux quantum @xmath19 .
the shift of index @xmath81 corresponds to the spectral flow @xmath82 , @xmath42 , discussed in the previous section .
altogether the single edge is described by a multiplet of partition functions @xmath59 , that is not modular invariant , meaning that the chiral anomaly , i.e. the spectral flow , implies a discrete gravitational anomaly .
more precisely , the @xmath59 transform in a linear unitary vector representation of @xmath83 , a subgroup of @xmath84 including only @xmath77 transformations and thus allowing fermionic statistics @xcite .
the dimension @xmath43 of the representation is equal to the value of the wen topological order @xcite .
the partition function for the quantum spin hall system made by a pair of laughlin states is obtained by combining chiral and antichiral sectors for up and down spins , respectively , thus obtaining the expressions @xmath85 . in the case of the hall effect of charge and spin ,
the presence of extended states in the bulk of the annulus allows for the matching of fractional charges between opposite edges . in the case of topological insulators , the bulk is insulating , thus the fractional charges should be matched locally at each edge ( the @xmath78 condition ) .
we thus obtain the following expression for the partition function of a single edge : z^ns(,)= _ = 1^p k^_^_- .
[ ns - pf ] this is invariant under @xmath86 .
it turns out that this quantity is formally equal to the quantum hall effect partition function for the system of two edges , i.e. for the whole annulus @xcite .
however , the physical interpretation in the case of topological insulators is rather different , since it only describes a single edge of the annulus and the different charge sectors do not correspond to bulk anyonic excitations , but just describe degenerate ground states at one edge .
the expression @xmath87 is not completely modular invariant , because the transformations in the quotient @xmath88 deform it into other quantities according to a general pattern .
indeed , the partition functions of fermionic systems always involve four terms corresponding to the four spins structures needed for defining spinors on the torus @xcite . these amount to choosing periodic @xmath89 and antiperiodic @xmath90 boundary conditions for fermion fields in each direction ( in general there are @xmath91 terms on a genus @xmath92 surface ) .
these terms are known as the neveu - schwarz @xmath93 and ramond @xmath94 sectors and their tildes , as follows : ns , , r , , ( aa ) , ( ap ) , ( pa ) , ( pp ) .
[ spin.struct . ] the expression ( [ ns - pf ] ) is identified with the neveu - schwarz sector since the natural fermionic boundary conditions are antiperiodic : z^ns=_a .
[ pf - trace ] the other expressions are defined as : z^r&=&_p , z^&=&_a , z^&=&_p , [ pf2-trace ] where periodic conditions in time introduce the sign @xmath95 . the modular transformations among the four terms are depicted in fig . [ fig6](b ) and are explicitly checked in the appendix .
they form a triplet , @xmath96 , and a singlet , @xmath97 .
each one of the four spin sectors is made of @xmath43 `` anyonic '' sectors : z^s=_=1^p k^s_^s_- , s = ns , , r , .
[ pfc1 ] furthermore , these @xmath98 are invariant under @xmath77 , i.e. possesses fermions excitations , under @xmath78 for charge matching and under @xmath79 for the spectral flow among the @xmath43 anyon sectors . regarding the @xmath99 sector ,
the partition function and its anyon sectors are defined as follows , z^(,)&= & z^ns ( + 1 , ) , k^_(,)&= & e^i_ k_(+1 , ) .
[ def_nst ] in this expression , the phase @xmath100 is included for convenience ; we also write @xmath101 and omit spin arrows for simplicity . note that the term brought by the @xmath13 transformation in the summation over fermions inside @xmath102 , i.e. @xmath103 , is proportional to @xmath104 owing to the half integer conformal dimension of fermions .
the ramond sector is similarly obtained by acting with @xmath105 on @xmath87 and @xmath97 is defined by inserting the @xmath104 sign into the ramond expression .
explicit examples will be given later and are collected in the appendix .
altogether , the pattern in fig . [ fig6](b ) shows the response of the topological insulator edge to modular transformations , i.e discrete diffeomorphisms not connected to the identity .
the four spin sectors , each one made of @xmath43 anyon sectors , realize a kind of decoupling of anyonic and fermionic properties , as it will be more clear in the following .
flux insertions and ( b ) modular transformations on the four spin sectors @xmath106 ( odd @xmath43 case).,width=566 ] we now consider the response to adding magnetic fluxes and recover the fu - kane - mele stability analysis in the context of partition functions . as already said , each spin sector is invariant under @xmath79 , the addition of one flux quantum .
the addition of half flux @xmath107 transforms the sectors as shown in fig . [
fig6](a ) .
in particular , the neveu - schwarz sector is mapped into the ramond sector , v^ : z^ns ( , ) z^ns(,+ ) = z^r ( , ) .
[ nstor ] the addition of any half integer number of fluxes @xmath108 yields the same result , up to a reshuffling of anyon sectors within @xmath109 ( cf .
eq.([pfc1 ] ) ) . in order to disentangle the anyonic degeneracy from the electron degeneracy relevant for the fu - kane - mele stability analysis @xcite , we follow th extention of the argument due to levin - stern @xcite .
we observe that the addition of @xmath43 fluxes creates an electron excitation within the same anyon sector , as it corresponds to a symmetry of each @xmath59 , v^p : k_ k_+p = k_ , q^==1 .
[ p.fluxes ] therefore , the addition of @xmath110 fluxes will create a spin one - half excitation in the topological insulator edge , @xmath111 , while staying in the same anyonic sector .
it is therefore convenient to define the ramond sectors by the action of @xmath112 : v^ : k_ ( , ) k_(,+)~k_+(,)= k^r_ ( , ) , [ def_r ] where @xmath113 stands for equality up to a global phase .
we can now extend the stability analysis discussed in the free fermion case ( section two ) . upon applying @xmath110 fluxes ,
the neveu - schwarz ground state @xmath114 , the lowest state in @xmath115 , evolves in the ramond ground state @xmath116 present in @xmath117 .
this becomes a spin - half partner of a degenerate kramers pair : following the argument of section two , this proves the stability of the topological phase in the system made by a pair of laughlin states , for any @xmath43 value . the existence of the kramers pair and the behaviour of the spectrum shown in fig .
[ fig4 ] can be easily checked by inspecting the low lying states contained in @xmath118 and @xmath119 : one expands to lowest order in @xmath120 , with @xmath121 , and checks the terms to @xmath122 , @xmath123 , i.e. not involving additional particles .
we can also recompute the spin parity of the neveu - schwarz and ramond ground states : ( -1)^2s_ns= _ ns ( -1)^2s_r = - _ r , s = q^= .
[ sp.groundst ] the validity of the levin - stern stability index discussed in the introduction ( eq.([ls - index ] ) ) is thus verified for this system : 2s=2q^= = 1 , ( -1)^2s=-1 , [ ls - index2 ] where @xmath124 and @xmath125 is the number of charge sectors , i.e. the periodicity of @xmath59 . the ( would - be ) spin transport @xmath8 involved in this index , equal to the hall current of one chiral component , is that relative to the addition of half of the fluxes needed for creating an electron excitation within a given anyon sector . in conclusion
, we have found the important fact that the spin parity of the ramond ground state is different from that of the neveu - schwarz ground state .
this is a manifestation of the discrete anomaly @xmath0 : different sectors of the path integral ( eq . and ) have associated different quantum numbers @xcite ; in the same way , the chiral sectors @xmath59 have associated different charges due to the @xmath12 anomaly ( eq.([v_transf ] ) ) . in a fermionic non - chiral system composed of the four spin sectors ( [ pf - trace]),([pf2-trace ] )
it is always possible to find a modular invariant partition function by summing over all sectors , z_ising= z^ns+z^+ z^r+z^. [ ising ] this is the so - called ising projection because it occurs in cfts applied to statistical models like the ising model , its supersymmetric generalizations etc . @xcite .
this quantity is @xmath126 invariant .
nonetheless , we would like to argue that the theory defined by @xmath127 may not be consistent with the tr symmetry of topological insulators , implying spin parity conservation . in presence of the @xmath0 anomaly ,
the partition function ( [ ising ] ) sums spin sectors with different values of the ground state spin parity , thus violating tr symmetry .
therefore , this symmetry is explicitly broken or not defined in that theory .
if we insist on preserving tr symmetry , we should not sum over spin sectors and leave them as the components of a four - dimensional vector , z_tr= ( z^ns , z^ , z^r , z^ ) , [ ztr ] that carries a non - trivial representation of @xmath128 .
therefore , the @xmath0 anomaly is associated to a @xmath129 gravitational anomaly . in the theory described by this set of partition functions ,
@xmath87 represent the tr invariant edge system , while the other functions , @xmath130 , are excited states of the system in presence of electromagnetic or gravitational backgrounds .
we thus obtain the following result , later shown to be valid in general : & & tr symmetry + anomaly no modular invariance topological insulator , & & tr symmetry + modular invariance no anomaly trivial insulator .
[ dog ] let us add some remarks : \i ) as in the case of the quantum hall effect , the anomaly can be cancelled globally on the whole system by combining the partition functions of the two edges of the annulus , leading to a global modular invariant .
\ii ) in the quantum hall state , the chiral partition functions like @xmath59 can not be combined into a modular invariant for a single edge ( unless a special case with @xmath131 ) . in the topological insulator , they can or can not be combined depending on the fate of the @xmath0 symmetry ; this is a manifestation of the symmetry protection of the topological phase , because this can became trivial if the symmetry is not enforced .
a related statement is that the stability of these systems is not solely determined by cft properties but by the way @xmath132-dimensional bulk symmetries are attached to cft edge states .
\iii ) in the @xmath18 laughlin spin state considered in this section , the two partition functions @xmath133 and @xmath134 can be obtained by canonical quantization of the following lagrangians .
@xmath135 is obtained by independent quantization of two copies of the chiral boson theory @xcite , s_1=_x(_t - v _ x ) + _ xc(_t+v _ xc ) , [ ] with rational compactified radius @xmath136 , @xmath137 , and antiperiodic boundary conditions .
the spectrum of virasoro states is given by @xmath138 in one chiral theory and @xmath139 in the other one , with @xmath140 .
the partition function @xmath134 is obtained by the quantization of the non - chiral compactified boson @xcite , s_2=(_t)^2 -v^2 ( _ x)^2 . [ ] one obtains the standard spectrum @xmath141 , with @xmath142 , that spans a two - dimensional even self - dual lorentzian lattice , characteristic of compactified bosonic theories @xcite @xcite .
the partition function resulting by summing over the lattice is modular invariant and can be rewritten in the form ( [ ising ] ) : z_ising= _ n , m^2 q^(+mr)^2 ^(+mr)^2 . [ selfdual ] of course , the choice of other boundary conditions in the theory @xmath143 leads to the other three sectors , @xmath130 .
the ising projection then corresponds to imposing the condition @xmath144 that eliminates fermionic excitations from both the @xmath89 and @xmath90 spectrum .
namely , @xmath134 does not describe fermionic excitations and for this reason it contains less states than @xmath87 .
on the other hand , the @xmath134 possesses additional states of the ramond spectrum . in conclusion ,
the stable topological system described by @xmath87 contains single fermion excitations and can be derived from @xmath143 ; the unstable , non tr symmetric phase has only bosonic states and is naturally described by @xmath145 .
\iv ) another indication that tr symmetry is not present in @xmath146 is given by the violation of the spin - statistics relation in the ramond sector . since @xmath109 and @xmath97 are @xmath13 invariant , they only contain states with integer conformal dimension , i.e. bosonic statistics . on the other hand ,
their spin is half integer , upon summing @xmath147 for ramond ground state to those of excitations present in @xmath148 , i.e. @xmath149 .
therefore , the spectrum does not respect spin - statistics . on the other hand , the partition function @xmath150 would obey the charge - statistics relation and
also be modular invariant , but it is not invariant under @xmath151 .
the conformal theories of general quantum hall edge states possess not only charged excitations but also neutral modes that can be abelian or non - abelian .
these theories have the affine symmetry @xmath152 , where @xmath60 is the charge symmetry and @xmath153 is another ( non - abelian ) symmetry characterizing the neutral part ( possibly a coset @xmath154 ) . the electron field in this theory
is represented by the product of a chiral vertex operator for the charge part and a chiral neutral field @xmath155 of @xmath154 : _ e = e^i _ e .
[ el - field ] the field @xmath155 should also have abelian fusion rules with all fields in the theory : this property is needed for the electrons to have integer statistics with all excitations and for the ground state wavefunction to be unique , being written as a cft correlator .
the field @xmath155 , called a simple current in the cft literature @xcite , can be used to build a modular invariant that couples neutral and charged parts non - trivially and fulfills the physical conditions on charge and statistics of the edge spectrum .
the general expression of the partition function for the hall edge states obtained in this way is determined uniquely by two inputs : the choice of neutral @xmath154 theory and of the abelian field @xmath155 that represents the electron neutral part .
these simple - current modular invariant partition functions were shown to reproduce earlier results obtained by physical insight in many models and to build new ones @xcite .
the construction starts from the one - edge partition sum for an anyon sector , generalizing the @xmath59 of the @xmath18 theory introduced in section 3.1 ( eq.([pf ] ) ) : this involves again a basic anyon plus any number of electrons added to it , with charge @xmath49 , @xmath156 .
it is characterized by @xmath157 , and the neutral quantum numbers @xmath158 . in terms of states and fields ,
the electron excitations are obtained by fusing the basic anyon field with many electron fields , always getting a unique output owing to the abelian fusion .
this conserves the charge @xmath157 and another neutral additive number @xmath159 of the simple current ( assume @xmath159 modulo @xmath160 for simplicity , i.e. a @xmath161 neutral charge ) .
such partition function takes the form @xcite : _
( , k ; kp ) ^_+a p mod k(,0 ) .
[ general.pf ] the @xmath162 are the characters for the charge part , while the @xmath163 are sums of @xmath154 characters for the neutral part , that are labelled by the abelian number @xmath159 and other , possibly non - abelian , quantum numbers collectively denoted by @xmath164 .
the explicit form of the neutral characters @xmath165 is not needed , only their symmetries and modular transformations are relevant .
equation ( [ general.pf ] ) can be explained as follows .
the basic anyon has quantum numbers @xmath166 , with @xmath159 modulo @xmath160 and @xmath157 modulo @xmath167 owing to the periodicity : k_(,k;kp)=k_+kp(,k;kp ) .
[ pk - period ] after adding one electron , the quantum numbers changes into @xmath168 ; then , after adding @xmath160 electrons these numbers return to those of the basic anyon .
this explains the @xmath160 terms in the sum ( [ general.pf ] ) .
the difference with respect to the @xmath18 case ( [ pf ] ) is that @xmath169-electron states couple to different neutral parts for @xmath169 modulo @xmath160 ; actually , each @xmath162 in ( [ general.pf ] ) only sums electrons with @xmath170 , owing to its different charge normalization .
another way to state this fact is that physical @xmath157 and neutral @xmath159 charges are related by a @xmath161 parity rule @xcite .
we can use pairs of these edge theories to model general interacting topological insulators .
the functions @xmath171 enjoy similar properties under modular transformations as the @xmath59 of section 3.1 and the partition function @xmath87 , can be written accordingly , that couple the up / down spin modes at one edge , though the @xmath78 charge condition ( [ u_transf ] ) : z^ns=_ , _ ^_-^. [ ns.gen.th ] in this sum , the range for the @xmath172 values is given by the wen topological order . in earlier works @xcite@xcite@xcite@xcite , we showed that known expressions for the partition functions of multicomponent abelian theories in the @xmath3-matrix formalism can be recast in the form of ( [ general.pf ] ) , ( [ ns.gen.th ] ) ; non - abelian states are also written in this form , as e.g. the read - rezayi parafermion states with @xmath173 .
analogous expressions are obtained for the other spin sectors @xmath174 , by acting with the @xmath13 and @xmath14 transformations as explained before : these partition functions take the same form ( [ general.pf ] ) with signs for the @xmath175 weight and slightly different @xmath161 pairing in the ramond sectors .
examples will be given later in this section and in the appendix .
the charge part @xmath59 of the sectors @xmath176 in ( [ general.pf ] ) is parameterized by two independent numbers @xmath177 , whose meaning can be understood from the expression ( [ pf ] ) : \i ) the values of the fractional charge are @xmath178 , @xmath179 , and the minimal charge is equal to @xmath180 .
\ii ) the hall current ( spin current ) is obtained by applying the @xmath79 transformation on ( [ general.pf ] ) , that acts on the charge part @xmath59 , causing the shift of quantum numbers : v : + , +k , q^=^= , [ k - hallcurr ] while the neutral characters in ( [ general.pf ] ) are not affected .
we see that @xmath181 is parameterized by the ratio of the numbers @xmath177 , but these might have common factors that are relevant in the stability argument . as in section 3
, we should find the number of fluxes that creates an electron excitation in the same anyon sector ( fractional charge ) and same neutral sector ( neutral quantum numbers ) , such that all fusion rules stay unchanged @xcite : owing to the periodicity of @xmath59 in ( [ pk - period ] ) , this number is given by @xmath43 .
then , the fu - kane - mele flux argument considers the change in spin parity due to adding @xmath110 fluxes .
this is given by , v^ : s = q^= ^= . [ phalf - fluxes ] therefore , the levin - stern index ( [ ls - index2 ] ) for the spin parity anomaly is : 2s= = k. ( -1)^2s=(-1)^k .
[ ls - index - k ] the stability analysis then continues by observing that for odd values of @xmath160 , the action of @xmath112 creates a kramers pair at the edge that is protected by tr symmetry ; then , this can not be gapped and the topological insulator is stable .
the action on the anyon sectors ( [ general.pf ] ) is , v^ : _ ^ ( , ) _ a=1^k k_+a p+kp/2 ( , k ; kp ) ^_+a p mod k(,0 ) ~_^ ( , ) , [ act.pf ] where the values of @xmath182 depends on the specific theory considered through the symmetries of its characters .
looking at the expressions ( [ general.pf]),([act.pf ] ) , it is clear that the neutral characters @xmath183 do not enter in the stability argument ,
i.e. in the determination of the index ( [ ls - index - k ] ) ; only the charge parts are relevant .
thus , the result ( [ ls - index - k ] ) holds for both abelian and non - abelian edge theories of topological insulators .
in particular , the relevant parameters @xmath177 are independent of the value of wen topological order . in conclusion
, we have extended the levin - stern stability criterion to any interacting topological insulator with time reversal symmetry . in the case of multicomponent abelian theories , the works @xcite@xcite@xcite found the explicit edge interactions that gap the system in the unstable cases , thus double checking the result of the flux argument .
such analysis of gapping interactions is not yet available for general non - abelian theories ; however , we can provide the following argument .
some well - known non - abelian states have been described as projections of so - called `` parent '' abelian states @xcite : for example , the @xmath184 state of distinguishable electrons and its @xmath160-component generalizations are parents of the pfaffian and @xmath161 parafermion hall states , respectively .
the non - abelian theories are obtained by projecting the abelian theories to states that have identical electrons .
since this projection does not affect the tr invariance of states , it commutes with the analysis of tr - invariant interactions in the abelian theory and extends it to these non - abelian theories .
in particular , we shall see later that the topological insulator made by pfaffian states is unstable . clearly , the value of the levin - stern index is equal in the unprojected ( abelian ) and projected ( non - abelian ) theories
. the stability of general topological insulators , corresponding to the @xmath0 spin parity anomaly , is again accompanied by modular non - invariance of the partition function . however , electromagnetic and gravitational responses are not always equivalent as in the @xmath18 case ( neutral modes are clearly sensible to coordinate changes but not to flux additions )
. we should distinguish the following cases , according to the parities of @xmath177 : \i ) for @xmath43 odd , the action of @xmath112 is not a symmetry of each spin sector and maps them one into another .
the transformations between neveu - schwarz and ramond sectors and among their tildes are the same as those of the @xmath18 theory ( see fig . [
fig6](a ) and eq.([def_r ] ) ) .
the anyon sector @xmath185 containing the @xmath186 ground state is naturally mapped into @xmath187 including the ramond ground state .
the modular invariant and non - invariant partition functions are as in the @xmath18 case : z_ising&=&z^ns + z^ + z^r + z^ , k even , unstable , z_tr&=&(z^ns , z^ , z^r , z^ ) , k odd , stable .
[ non-inv.pf ] \ii ) for @xmath43 even , the action of @xmath112 maps each spin sector into itself and thus differs from the modular transformations ( see fig . [ fig7 ] ) . for @xmath160 odd
, the @xmath9 anomaly manifests itself within each spin sector , as a difference in spin parity between the ground state and another `` anyon '' ground state ( actually degenerate ) .
the tr symmetry of the theory then requires to splitting each spin sector in two subsectors , @xmath188 , @xmath189 , that are related by @xmath112 : @xmath190 and collect anyon sectors of same spin parity .
these subsectors carry a eight - dimensional representation of the modular group , namely the associated gravitational anomaly is slightly stronger . finally , for @xmath160 and @xmath43 both even , there is no anomaly and the @xmath134 partition function is consistent with tr symmetry . summarizing , in all cases modular non - invariance
is associated to stability and the @xmath0 anomaly .
flux insertions and ( b ) modular transformations on the four spin sectors @xmath106 ( even @xmath43 case).,width=566 ] the jain states are the prominent multicomponent abelian states in the fractional quantum hall effect . for @xmath160 components ,
their @xmath191 cft involves a lattice of conformal dimensions ( specified by the so - called @xmath3 matrix ) in which the abelian symmetry @xmath192 is enhanced to @xmath193 @xcite .
the fusion rules are abelian but there are manifestations of the @xmath194 symmetry and its center @xmath161 .
consider the topological insulator made by a pair of spin up and spin down jain states ; the filling fraction , minimal charge and stability index are : ^= , e^*= , , ( -1)^2s=(-1)^k , [ jain - qn ] showing that the system is stable ( unstable ) for @xmath160 odd ( even ) . the jain partition function in the @xmath186 sector has been studied extensively @xcite .
the previous formulas identify the parameters entering the stability analysis are @xmath160 and @xmath195 ; they have no common factor , @xmath196 , and @xmath43 is always odd .
moreover , the topological order is equal to @xmath43 .
thus , this case is very similar to that of laughlin states , and the pattern of flux insertions and modular transformations among the four spin sectors is the same as that already discussed in section 3 ( fig .
[ fig6 ] ) .
on the other hand , there are @xmath197 neutral modes and the partition functions have the simple - current form discussed in the previous section .
the anyon sectors in the @xmath186 sector take the form eq.([general.pf ] ) @xcite : ^ns_(,)=_a=0^k-1 k_+ap(,k;kp ) _ +ap mod k(,0 ) , =1, ,kp .
[ jainns - pf ] in this expression , @xmath198 are the characters of the @xmath199 affine algebra with abelian label @xmath44 modulo @xmath160 @xcite .
it is apparent that charged and neutral sectors are paired by the @xmath161 section rule @xmath200 modulo @xmath160 , whose origin can be understood as follows @xcite@xcite .
the abelian @xmath160-dimensional lattice of conformal dimensions specified by the @xmath3 gram matrix is not orthogonal , i.e. the charged and neutral excitations are not independently generated .
the decomposition underlying the expression ( [ jainns - pf ] ) is obtained by embedding the abelian lattice into a thinner one that is orthogonal in the charge and neutral directions . in this last lattice ,
the `` physical '' points are those obeying the @xmath161 selection rule .
the anyon sectors are periodic by @xmath201 , due to @xmath202 modulo @xmath160 ; thus there are @xmath43 independent ones ( the value of the topological order ) , that can be chosen to be @xmath203 , @xmath204 .
let us present more explicit formulas in the @xmath205 case ( unstable ) ; the partition functions of the four spin sectors are : [ jain - totpf ] ^ns_2a ( , ) & = & k_2a _ 0 + k_2a+p _ 1,a=1, ,p , ^_2a ( , ) & = & k_2a _ 0 - k_2a+p _ 1 , ^r_2a ( , ) & = & k_2a _ 1 + k_2a+p _ 0 , ^_2a ( , ) & = & -k_2a _ 1 + k_2a+p _ 0 , [ jain2-pf ] where @xmath206 and @xmath207 .
note that first ( second ) @xmath59 in each expression includes an odd ( even ) number of fermion excitations ; then , the relative sign appearing in the @xmath208 and @xmath209 expressions accounts for the @xmath104 factor in the trace ( [ pf2-trace ] ) .
note also that the @xmath0 parity rule between charge and neutral quantum numbers that is different in the @xmath186 and @xmath30 sectors .
the transformation properties of these sectors are given in appendix a. the partition functions of the four sectors are : z^()=_a=1^p ^()_2a , = ns,,r , .
[ jain2-totpf ] the partition function for this unstable system is given by @xmath134 .
we briefly discuss the general abelian edge theories that have been studied extensively in @xcite@xcite@xcite and show how they fit in the present analysis .
abelian conformal theories with central charge @xmath210 are characterized by a @xmath211-dimensional lattice of conformal weights and charges , that specifies the statistical phases @xmath212 , the electric charge @xmath213 and spin @xmath14 of excitations through the following formulae , = _ i , j=1 ^ 2n*n*_i*k*^-1_ij*n*_j , q=_i , j=1 ^ 2n*t*_i*k*^-1_ij*n*_j , s=_i , j=1 ^
2n*s*_i*k*^-1_ij*n*_j , _ i ^2n .
[ ab - spec ] these involve the gram matrix @xmath214 of the lattice , the charge and spin vectors , @xmath215 , @xmath216 , and the integer vector @xmath217 specifying the excitations . in particular , excitations with integer statistics , i.e. electrons and their compounds , are described by the dual lattice of @xmath218 .
the gram matrix is expressed by the basis vectors @xmath219 of the lattice as @xmath220 , where @xmath221 , @xmath222 is the signature of the lorentzian metric , expressing the chirality of excitations @xcite@xcite .
basis vectors @xmath223 and rescaled eigenvectors @xmath224 form equal matrices , up to transposition . in time reversal
invariant systems , there is an equal number of positive and negative chiralities .
next , we can choose a basis in which the first ( second ) @xmath169 components describe the spin - up ( down ) modes .
then the action of the tr transformation @xmath225 and the other quantities take a two - dimensional block form : =( cc 0 & 1 + 1 & 0 ) , =( c t + t ) , =( c s + -s ) , =( c n^ + n^ ) .
[ two - block ] the tr symmetry of the spectrum implies the condition @xmath226 , with solution : =( cc k & w + w^t & -k ) , k^t = k , w^t =- w .
[ gen - k ] it follows that @xmath218 has eigenvectors in pairs @xmath227 , that are integer valued @xcite@xcite@xcite .
the case of jain states considered before corresponds to @xmath228 , and , moreover , to a positive definite metric @xmath3 , because excitations of same spin have same chirality : =( cc k & 0 + 0 & -k ) , n^_i k_ij n^_j 0 .
[ block - k ] the partition functions of topological insulators made of these abelian theories can be obtained within the @xmath3-matrix formalism , and the expressions for anyon and spin sectors are given in appendix a. the decomposition of anyon sectors in simple - current form ( [ general.pf ] ) , schematically @xmath229 , is obtained by separating the lattice into charged and neutral sublattices , whose points are coupled by a parity rule as explained before .
next we discuss the systems with neutral modes counter - propagating with respect to the charged mode of the same spin type .
these are described by the block - diagonal form ( [ block - k ] ) , where @xmath3 has mixed signature , i.e. is not positive definite : an example is given by the jain states with @xmath230 .
as discussed in @xcite , the simple current expression ( [ general.pf ] ) of partition functions is still valid with the replacement @xmath231 for the opposite neutral chirality , schematically @xmath232 .
finally , the case @xmath233 describes interacting excitations with up and down spins .
indeed , let us choose the standard basis specified by @xmath234 and @xmath235 ; the statistical phases are computed from @xmath236 , where the @xmath237-th electron is represented by @xmath238 .
then , the correlation of up and down electrons , @xmath239 , has a phase proportional to @xmath240 , for @xmath241 and @xmath242 .
this is due to the coupling of the charge mode , e.g. spin up , with neutral modes of opposite spin . in this case
, the simple current decomposition of anyon sectors ( [ general.pf ] ) applies in the form @xmath243 . in conclusion ,
the simple current form of the partition function is valid in all abelian cases ; the stability argument discussed earlier applies equally , because it does not involve the properties of neutral modes .
let us finish the discussion of abelian theories by considering the example of type ( [ block - k ] ) with two - component @xmath3 matrix , k= ( cc 3 & 1 + 1 & 5 ) , [ k-3115 ] and charge vector @xmath244 . using previous formulas , we obtain the charge spectrum , minimal charge and filling fraction of the corresponding hall state : q= , n_1,n_2 , e^*= , ^= , [ qn-3115 ] while the topological order is @xmath245 .
next we consider the topological insulator made by a pair of these chiral states .
the parameters entering the stability analysis are @xmath246 , thus the topological state is stable , 2s==3 , ( -1)^2s=-1 .
[ two - comp ] this example shows that the stability is independent of the number of fermion modes and of the value of the topological order .
the pfaffian state is the simplest example of non - abelian quantum hall states @xcite .
following the previous discussion of stability , it can be analyzed together with its parent abelian state , the so - called @xmath184 which has the same charge spectrum and filling fraction ( see ref .
@xcite for a complete analysis of this relation ) . from the matrix @xmath247
, we find : ^=,e^*=,2s= 2 , ( -1)^2s=1 . [ qn-331 ] thus the topological insulators made by pairs of these hall states are unstable .
the parameters entering the stability analysis are @xmath248 .
let us first discuss the abelian state .
its topological order is @xmath249 , thus there are @xmath17 anyon sectors and a @xmath0 parity rule . in the @xmath186 sector , they read : _
= k__+k_+4_+2,=1, ,8 , [ 331-nsanyon ] where @xmath250 and @xmath251 .
the neveu - schwarz partition function reads @xcite : z^ns_(331)=_a=0 ^ 3 k_a_a + k_a+4 _
a+2 ^2 + k_a_a+2 + k_a+4 _ a ^2 .
[ 331-nspf ] note that the four charge sectors appears twice in the spectrum coupled to different neutral parts .
the expressions for the other spin sectors are given in appendix .
in the pfaffian state , the neutral part is provided by the characters of the ising model , i.e. the @xmath9 parafermions , @xmath252 , labelled by @xmath253 modulo @xmath254 and @xmath255 , non - vanishing for @xmath256 modulo @xmath257 , and obeying @xmath258 @xcite .
there are only three independent values , @xmath259 , @xmath260 and @xmath261 , denoted as the corresponding conformal fields .
the anyon sectors are ( @xmath186 sector ) : _ a^=k_a_a^+k_a+4_a+2^ , a=0,1,2,3,=0,1,2,a= mod 2 .
[ pfaffian - anyon ] it is apparent that the charge parts in the @xmath184 and pfaffian states are equal , while the neutral parts of the latter obey some symmetries that reduce the topological order from @xmath17 to @xmath262 ; one indeed checks that @xmath263 and @xmath264 @xcite . the @xmath186 partition function reads @xcite : z^ns_pf&=&_a=0,2 k_ac^0_a + k_a+4 c^0_a+2 ^2 + k_ac^0_a+2 + k_a+4 c^0_a ^2 + & & ( k_1+k_-3 ) _ 1 ^ 1 ^2 + ( k_3+k_-1 ) _ 1 ^ 1 ^2 , [ pfaffian - ns ] where charge sectors appear again twice , but are coupled differently to neutral states . in particular , the two @xmath184 sectors with neutral charge @xmath265 ( resp .
@xmath266 ) are projected to a single one .
this example clarifies that the fu - kane - mele stability argument is the same in both theories , since it only deals with the charge parts .
the expressions of the other spin sectors for the pfaffian topological state are also given in the appendix .
these unstable theories are characterized by @xmath177 both even , thus flux insertions and modular transformations act differently on the four spin sectors ( fig .
[ fig7 ] ) . as discussed before
, @xmath267 is an isometry of each spin sector .
since there is no associated @xmath0 anomaly , the modular invariant @xmath134 partition function is consistent with tr symmetry .
the read - rezayi states @xcite are generalization of the pfaffian state involving neutral modes of the @xmath161 parafermions that can be described by the coset @xmath268 @xcite .
the quantities entering in the stability index ( [ ls - index ] ) are ^= , e^*= , 2s = k , ( -1)^2s=(-1)^k , [ rr - qn ] where @xmath269 and @xmath270 . in this case ,
@xmath271 , thus the topological insulators made by a pair of these states is stable ( unstable ) for @xmath160 odd ( even ) .
note that @xmath160 and @xmath43 have the same parity : \i ) for @xmath160 and @xmath43 odd , these numbers have no common factor , and the flux insertions and modular transformations follow the pattern of the stable , odd @xmath160 jain states and of the @xmath18 theory ( fig .
[ fig6 ] ) .
the modular non - invariant partition function takes the form @xmath133 in ( [ non-inv.pf ] ) .
\ii ) for @xmath160 and @xmath43 even , they have a common factor of @xmath257 and the flux insertions and modular transformations are the same as those of the @xmath205 case discussed in the previous paragraph ( fig . [ fig7 ] ) .
the modular invariant partition function is @xmath134 .
all partition functions and modular transformations are described in appendix .
let us briefly discuss the form of the neveu - schwarz anyon sectors , taken from ref.@xcite .
these read : ^_a = _
b=1^k k_a+bp(,k;kp ) c^_a+2b(,0,2k ) , a= mod 2,p=2+km .
[ rr - nsanyon ] the charge characters @xmath59 with periodicity @xmath167 , are coupled to the @xmath161 parafermion characters @xmath272 , that are specified by the @xmath273 quantum number @xmath274 , and the abelian number @xmath159 modulo @xmath275 .
the @xmath161 parity rule between the two abelian numbers is @xmath276 modulo @xmath160 ( note that @xmath277 modulo @xmath160 ) .
the parafermion characters obey the periodicities @xmath278 and vanish for @xmath279 modulo @xmath257 . taking into account these properties ,
one finds the periodicity @xmath280 , implying @xmath281 independent anyon sectors , the topological order .
in this paper we have obtained the general form of the partition function of edge excitations in non - chiral topological states protected by time - reversal symmetry , such as the quantum spin hall effect and topological insulators .
the study of partition functions has allowed us to discuss the flux argument for the stability of the topological phase in great generality and to extend it to interacting systems possessing non - abelian edge excitations .
we have emphasized the anomaly in the @xmath0 edge spin parity that is associated to stable topological phases .
we considered the modular transformations of the partition function that map the four spin sectors , neveu - schwarz , ramond and their tildes , among themselves .
we have found that the @xmath0 anomaly is accompanied by modular non - invariance , that is a kind of discrete gravitational anomaly .
we also discussed the cases in which the electromagnetic and gravitational responses are equivalent and when they are different . among the possible directions of future progress , we mention the extension of our stability analysis to topological superconductors @xcite .
let us briefly discuss this point .
from the point of view of edge dynamics , topological superconductors amount to systems of ( interacting ) @xmath15 neutral chiral majorana modes with spin up and @xmath15 ones with opposite spin and chirality , thus making @xmath15 ising cfts @xcite .
the topological phases are protected by the @xmath282 symmetry of independent edge spin parity for each chirality , @xmath283 and @xmath284 .
this follows from tr symmetry and the absence of spin - flip terms in the bulk hamiltonian @xcite .
since this symmetry does not allow any mass term , the non - interacting topological insulators are classified by the @xmath23 index equal to @xmath15 .
on the other hand , for @xmath285 a non - trivial quartic interaction was found that gaps the system without breaking the symmetry explicitly or spontaneously @xcite@xcite .
this interaction is possible because the ising cfts have spin @xmath286 and disorder @xmath58 fields of dimensions @xmath287 , such that the product of eight of them has dimension @xmath288 and can bosonized and refermionized to obtain a quartic fermionic term respecting the symmetry .
therefore , the classification of interacting topological superconductors is associated to a @xmath289 index ( at least ) . for even @xmath15
, the free majorana fermions have @xmath290 and yield the same fermion systems considered in this paper .
however , the neutral modes can not be probed by flux insertions for studying stability . in an interesting paper @xcite , ryu and
zhang suggested to consider the modular non - invariance as a test of stability of topological superconductors . in practice
, they associated the instability to the modular invariance of the partition function for the singlet sector of the theory with respect to the @xmath282 symmetry , @xmath291 , i.e. involving only bosonic chiral states .
indeed , this projected partition function is modular invariant for @xmath285 , as it corresponds to the so - called gliozzi - scherk - olive projection @xmath292 in superstring theory @xcite .
this results provided a symmetry argument for the instability of @xmath285 topological superconductors @xcite@xcite ( also extended to other systems in @xcite ) . in our analysis , we considered the partition function of the full theory @xmath134 that is always modular invariant , but argued that it might not be consistent with tr invariance , owing to anomalies in some of its sectors .
the two criteria of stability do not seem to be equivalent in general , but at least they agree on the instability of the @xmath285 case .
the argument goes as follows . in section 3
, we have seen that in free fermionic systems , the half flux insertion maps neveu - schwarz into ramond sectors , i.e. we have the equivalence @xmath293 , between electromagnetic and gravitational responses .
moreover , @xmath151 does the correct flux insertion for measuring the anomaly .
we then consider the chiral spin of the excitation in the @xmath30 sector obtained from the @xmath186 sector by the @xmath105 transformation : for @xmath294 fermion modes , this is @xmath295 . the first @xmath282 anomaly free case is for @xmath285 , with chiral indices : ( -1)^n_=(-1)^n_=1,n_f=8 .
[ 8fermions ] therefore , the modular invariant partition function @xmath134 is free of @xmath282 anomalies for @xmath285 , thus confirming the instability of this system in our analysis .
more general results can not be provided at present , owing to two main difficulties : the correct measure of spin anomaly in other systems and the form of the partition function to be associated to the edge of interacting topological superconductors . regarding the second point
, the absence of the @xmath78 condition , matching the charges of the two chiralities , allows many possible expressions for the edge partition function , of the form @xmath296 , with @xmath297 any symmetric positive integer matrix commuting with the @xmath14 and @xmath13 transformations @xcite .
another possible development concerns the comparison with the stability criterion proposed in @xcite , which is based on properties of fusion rules , again related to modular transformations through the verlinde formula .
finally , there is the study of stability of three dimensional systems : in particular , recent works @xcite suggest the existence of topological states at the @xmath298 surface of a @xmath299 topological insulators , whose stability could be different from that of strictly @xmath298 systems considered here .
* acknowledgments * ac and er thanks t. h. hansson , d. seminara , g. viola , p. wiegmann and g. r. zemba for interesting discussions .
this work was partially supported by the ec grant m. curie action irses-295234 `` quantum integrability , conformal field theory and topological quantum computation '' .
in the following we give the expressions of partition functions for the four spin sectors @xmath300 , that describe the edge excitations of topological insulator models examined in the main text .
we describe their behavior under flux insertion and modular transformations .
the laughlin states discussed in section 3 are described by the @xmath18 cft @xcite .
the anyon sectors for the chiral modes of the @xmath186 and @xmath99 spin sectors are , for @xmath301 , @xmath43 odd , @xcite : @xmath302 with @xmath303 $ ] is a non - holomorphic prefactor and @xmath63 the dedekind function [ dedekind ] ( ) = q^_n=1^(1-q^n ) , q=(i2 ) .
the anyon sectors of the @xmath30 and @xmath209 spin sectors are defined by : @xmath304 the edge partition functions for each spin sector ( eq . and ) are obtained by matching the fractional charge of the chiral and antichiral anyon sectors locally at the edge ( see @xmath78 condition ( [ u_transf ] ) ) , @xmath305 the transformation of the anyon sectors ( ( [ nschar1 ] ) and ( [ rchar1 ] ) ) under the modular group , generated by @xmath14 and @xmath13 , and for the insertion of one and @xmath110 fluxes through the annulus , the @xmath79 and @xmath112 transformations , respectively , are obtained by extending the calculations of ref .
@xcite @xcite .
altogether they are : * @xmath14 @xmath306 with @xmath307 * @xmath13 @xmath308 with [ t - element ] t_a= ( 2 i ( - ) ) , t_b= ( 2 i ( + ) ) .
* @xmath79 @xmath309 with [ v - element ] v__0(,)= ( -2 i ( re + re ) ) * @xmath112 @xmath310 with [ vp2-element ] v__0= ( -2 i ( re+ re ) ) . upon using these formulas
, we obtain that the transformations of partition functions ( [ pfc1-spinsector ] ) illustrated in fig.([fig6 ] ) .
as discussed in section 3.4 , the modular invariant partition function @xmath311 ( [ ising ] ) is not consistent with tr symmetry owing the presence of the @xmath9 anomaly , then the laughlin - type topological insulators are stable . the topological insulators with two - component jain edge systems discussed in section 4.3.1 , are described by the @xmath312 cft with symmetry @xmath313 ; the two parameters that enter the stability argument ( see section 4.2 ) are @xmath314 .
the characters of the @xmath315 charge part are expressed by the functions @xmath316 of the previous section ( eq.([pf ] ) ) , where @xmath317 and @xmath318 .
the @xmath319 neutral characters are : @xmath320 these have the periodicity @xmath321 and obey the modular transformations ( with @xmath205 ) @xcite @xmath322 then , the chiral anyon sectors for the four spin sectors are : @xmath323 for each spin sectors the edge partition functions are obtained by coupling the chiral and antichiral modes and summing over @xmath324 , as follows ( see eq.([jain2-totpf ] ) ) : z^()=_a=1^p ^()_2a ^()_-2a , = ns,,r , . [ jain - pf - tot ]
* @xmath14 @xmath325 with [ sjain - el ] s_a , a^= ( 2i ) , e^i =( i ( ) ) .
* @xmath13 @xmath326 with [ tjain - el ] t_a= ( 2 i ( - ) ) .
* @xmath79 @xmath327 with [ vjain - el ] v__0= ( -2 i ( re + re ) ) .
* @xmath112 @xmath328 with [ vp2jain - el ] v__0= ( -2i ( re+ re ) ) . using these transformations
, we obtain that the partition functions for the jain states ( [ jain - pf - tot ] ) transform as in the @xmath18 case under the @xmath112 and modular transformations , because the number of charge sectors @xmath43 is odd for the jain theory ( see fig.([fig6 ] ) and section 4.2.1 ) . owing to the absence of @xmath9 anomaly
, the modular invariant partition function @xmath311 is consistent with tr symmetry , then the topological phase is instable . in the following ,
we also give the partition functions for the general multicomponent jain theory with @xmath191 and symmetry @xmath329 .
the stability parameters are @xmath330 .
the form of the transformations depends on the parity of the @xmath160 parameter : \i ) for even @xmath160 , we remember from ref .
@xcite that the charge characters are given by @xmath331 , with @xmath332 and @xmath318 .
the neutral characters @xmath333 belong to the @xmath334 affine algebra and have the periodicity and modular transformations ( [ mod - neutro - jain ] ) . starting from the anyon partition function of the neveu - schwarz sector ( [ jainns - pf ] ) , we can act with the @xmath13 and @xmath105 modular transformations to obtain the other spin sectors . altogether
they are : @xmath335 note that for @xmath205 we obtain the expressions given before ( eq.([jain - anyon2 ] ) ) .
\ii ) for odd @xmath160 , the charge characters are different for each spin sector ( as in the @xmath18 case ) . for the @xmath186 and @xmath99 spin sectors they read : [ jain - kodd - ns ] k^ns_ka+p(,k ; kp)=k_ka+p(,k ; kp ) , the corresponding expressions for the @xmath30 and @xmath336 spin sectors are defined by @xmath337 upon combining them with the neutral characters @xmath338 of the @xmath334 affine algebra , the anyon sectors take the general form for the simple current modular invariants , as explained in section 4 : [ any - kodd - jain ] _
ka^()(,;k)=_=1^k k^()_ka+ p(,k;kp ) _
( , 0 ) , = ns , , r , .
the transformations under the modular group and the insertion fluxes are those shown in fig.[fig6 ] , since @xmath43 is always odd in the jain theory .
let us discuss the general abelian states whose spectrum of charge and statistics is described by the @xmath3-matrix formalism ( see section 4.3.2 ) . in the case
that the @xmath169 chiral ( spin - up ) and @xmath169 antichiral ( spin - down ) edge modes are independent , the @xmath218 matrix takes the block diagonal form ( [ block - k ] ) and the @xmath340 block @xmath3 is definite positive , integer valued and with @xmath341 odd , @xmath342 . in the basis in which the charge vector is @xmath343 , where @xmath344 @xmath345dimensional , the characters of the chiral modes have been written in ref.@xcite .
the chiral anyon sectors modes of the @xmath186 and @xmath99 spin sectors are ( with @xmath346 ) : @xmath347\biggr),\ ] ] @xmath348\biggr),\notag \end{split}\ ] ] with the prefactor @xmath349 ; those of the @xmath30 and @xmath336 sectors are related to these by @xmath350 in these equations , @xmath351 is a @xmath169-dimensional vector whose components are the diagonal elements of the @xmath3 matrix : @xmath352 the edge partition functions for each spin sectors are @xmath353 * @xmath14 @xmath354 with @xmath355 * @xmath13 @xmath356 with @xmath357 in order to obtain these transformations , we used the symmetry of the @xmath3 matrix and the odd parity of its diagonal elements , leading to : @xmath358 = \text{exp}\bigg [ 2 \pi i \frac{1}{2 } k_{ii}\ell^2_i \bigg]= ( -1)^{\ell^t b}.\ ] ] * @xmath79 @xmath359 with [ v - label - el ] v__0 ( , ) = ( -2 i t^t k^-1 t ( re + re ) ) . in this theory , the partition functions ( [ pf - abelian ] ) transform under the modular group as depicted in fig.([fig6])(b ) .
we have to more careful about the @xmath112 transformation : according to the discussion in section ( 4.2.1 ) , there are two different results depending on the parity of @xmath43
. this number can be obtained from the minimal value @xmath180 in the charge spectrum . in general
, we can write : @xmath360 with @xmath361 we distinguish the two cases : \i ) if @xmath43 is odd , the @xmath112 transformation maps @xmath186 sectors into @xmath30 sectors , up to a reshuffling of the anyon label , @xmath362 , as follows : [ vp2resch ] v^ : k^ns _ k^r_ , and similarly for the other sectors
. then the partition function of the neveu - schwarz sector is mapped in that of the ramond sector and so on , as depicted in fig.([fig6])(a ) . for the ( 331 ) state , we recall from the main text that @xmath248 , the charge character is @xmath364 , with @xmath365 , and the abelian neutral character is @xmath366 @xcite .
the anyon sectors for the chiral ( spin - up ) modes are : @xmath367 then , the edge partition functions for the neveu - schwarz and ramond spin sectors of the ( 331 ) topological states can be written as : @xmath368 the ( 331 ) anyon sectors ( [ 331-any - spin ] ) transform under the modular group and the insertion of fluxes according to the abelian formulas eq.([s - abel]-[vp2-abel ] ) with @xmath369 , and @xmath370 even .
the edge partition function for the pfaffian states in the neveu - schwarz spin sector is given in ( [ pfaffian - ns ] ) @xcite .
we obtain the partition functions of the other spin sectors acting with @xmath13 and @xmath105 on @xmath371 ( [ pfaffian - ns ] ) as described in section ( 3.2 ) .
they read : @xmath372 @xmath373 we have written the neutral characters in ( [ pfaffian - pf ] ) with the same symbol of the ising fields @xmath374 that model the neutral excitations of this system @xcite @xcite . the spin sectors partition functions of ( 331 ) and pfaffian topological states , eqs .
( [ 331-pf ] ) and ( [ pfaffian - pf ] ) , respectively , transform in the same manner under @xmath112 and the modular group because the two theories have the same @xmath177 parameters entering the stability analysis , as discussed in section ( 4.3.3 ) . since @xmath43 is even ,
these transformations are shown in fig.([fig7 ] ) .
the absence of the @xmath9 anomaly in both cases allows for the @xmath311 modular invariant ( [ ising ] ) to be tr symmetric , such that the two topological phases are unstable .
the read - rezayi models @xcite are based on the neutral @xmath375 parafermion conformal theories with central charge @xmath376 , described by the coset construction @xmath377 @xcite . for these theories
the values of the stability parameters are @xmath378 with @xmath379 and @xmath380 .
the structure of partition functions depends on the parity of @xmath160 .
the charge characters are given by the functions ( [ pf ] ) @xmath381 with periodicities @xmath382 .
the @xmath375 parafermionic characters that describe the neutral part are denoted by @xmath383 , and have the following periodicities and modular transformations @xcite , @xmath384 starting from the neveu - schwarz anyon sectors given in ref.@xcite and acting with @xmath13 and @xmath105 , we find those of the other spin sectors .
altogether they read : @xmath385 where @xmath386 , and @xmath387 .
the partition functions in the corresponding spin sectors are : @xmath388 we note that the partition functions of the pfaffian state ( eq.[pfaffian - pf ] ) are obtained by choosing @xmath389 and @xmath205 in the previous formulas .
the transformations of the anyon sectors ( [ rr - anyon ] ) and the partition functions ( [ rr - keven - pf ] ) under the insertion of fluxes and the modular group are the same of the pfaffian state , represented in fig.([fig7 ] )
. in this case the @xmath9 anomaly is absent , the modular invariant partition function @xmath311 ( [ ising ] ) is consistent with tr symmetry and the topological phases are unstable .
it is necessary to introduce different charge characters for each spin sector . for the @xmath186 and @xmath99 sectors
they are : @xmath390}.\notag \end{aligned}\ ] ] those of the @xmath30 and @xmath336 spin sectors are defined by : @xmath391 these charge characters have the same periodicities ( [ pk - period ] ) , i.e. @xmath392 . making use of @xmath375 parafermionic characters @xmath383 , the anyon sectors for four spin sector take the usual form of simple current invariants : @xmath393 the edge partition functions of the four spin sector have the same expression of those in the even @xmath160 case . because the values of @xmath177 are odd , as discussed in section ( 4.2.1 ) , it is easy to show that under the @xmath112 transformation and the modular group the anyon sectors and the partition functions transform as represented in fig.([fig6 ] ) .
owing to the presence of @xmath394 anomaly , the modular invariant partition function @xmath311 is inconsistent with tr symmetry , and these non - abelian topological phases are stable .
99 x. l. qi , s. c. zhang , `` topological insulators and superconductors '' rev .
* 83 * ( 2011 ) 1057 ; e. fradkin , field theories of condensed matter systems , 2nd edition , cambridge univ .
press ( 2013 ) , cambridge uk . m. knig , s. wiedmann , c. bruene , a. roth , h. buhmann , l. w. molenkamp , x. l. qi , s. c. zhang , `` quantum spin hall insulator state in hgte quantum wells '' , science * 318 * ( 2007 ) 766 ; m. koenig , h. buhmann , l. w. molenkamp , t. hughes , c. -x .
liu , x. -l .
qi , s. c. zhang , `` the quantum spin hall effect : theory and experiment '' , j. phys .
* 77 * ( 2008 ) 031007 .
d. hsieh , d. qian , l. wray , y. xia , y. s. hor , r. j. cava , m. z. hasan , a topological dirac insulator in a quantum spin hall phase " , nature * 452 * ( 2008 ) 970 ; y. xia , d. qian , d. hsieh , l. wray , a. pal , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , m. z. hasan ,
observation of a large - gap topological - insulator class with a single dirac cone on the surface " , nature physics * 5 * ( 2009 ) 398 ; y. l. chen , j. g. analytis , j .- h .
chu , z. k. liu , s .- k .
mo , x. l. qi , h. j. zhang , d. h. lu , x. dai , z. fang , s. zhang , i. r. fisher , z. hussain , z .- x .
shen , `` experimental realization of a three - dimensional topological insulator bi@xmath395te@xmath396 '' , science * 325 * ( 2009 ) 178 .
x. g. wen , quantum field theory of many - body systems , oxford univ .
press ( 2007 ) , oxford .
x. chen , z. x. liu , x. g. wen , `` two - dimensional symmetry - protected topological orders and their protected gapless edge excitation '' , phys .
b * 23 * ( 2011 ) 235141 . c. l. kane , e. j. mele , `` @xmath397 topological order and the quantum spin hall effect , '' phys .
* 95 * ( 2005 ) 146802 ; l. fu , c. l. kane , `` time reversal polarization and a @xmath397 adiabatic spin pump '' , phys .
b * 74 * ( 2006 ) 195312 , `` topological insulators with inversion symmetry '' , phys . rev .
b * 76 * ( 2007 ) 045302 ; l. fu , c. l. kane , e. j. mele , `` topological insulators in three dimensions '' , phys . rev . lett.*98 * ( 2007 ) 106803
. j. e. moore , l. balents , `` topological invariants of time - reversal - invariant band structures '' , phys .
b * 75 * ( 2007 ) 121306 ; r. roy , `` @xmath9 classification of quantum spin hall systems : an approach using time - reversal invariance '' , phys . rev .
b 79 ( 2009 ) 195321 .
m. levin , a. stern , `` fractional topological insulator '' , phys .
* 103 * ( 2009 ) 196803 , `` classification and analysis of two dimensional abelian fractional topological insulators , '' phys .
b * 86 * ( 2012 ) 115131 .
t. neupert , l. santos , s. ryu , c. chamon , c. mudry , `` fractional topological liquids with time - reversal symmetry and their lattice realization '' , phys .
b * 84 * ( 2011 ) 165107 , `` time - reversal symmetric hierarchy of fractional incompressible liquids '' , phys . rev .
b * 84 * ( 2011 ) 165138 . y. m. lu , a. vishwanath , `` theory and classification of interacting integer topological phases in two dimensions : a chern - simons approach '' , phys .
b * 86 * ( 2012 ) 125119 . p. di francesco , p. mathieu , d. snchal , conformal field theory , springer - verlag ( 1997 ) new york .
a. cappelli , g. r. zemba , `` modular invariant partition functions in the quantum hall effect , '' nucl .
b * 490 * ( 1997 ) 595 .
a. cappelli , l. s. georgiev , i. t. todorov , `` a unified conformal field theory description of paired quantum hall states , '' commun .
* 205 * ( 1999 ) 657 .
a. cappelli , g. viola , `` partition functions of non - abelian quantum hall states , '' j. phys .
a : * 44 * ( 2011 ) 075401 . j. polchinski , `` string theory .
2 : superstring theory and beyond , '' cambridge univ . press ( 1998 ) cambridge uk .
s. ryu , s. c. zhang , `` interacting topological phases and modular invariance , '' phys .
b * 85 * ( 2012 ) 245132 .
l. fidkowski , a. kitaev , `` effects of interactions on the topological classification of free fermion systems '' , phys .
b * 81 * ( 2010 ) 134509 . x. l. qi , `` a new class of ( 2 + 1)-dimensional topological superconductors with topological classification '' , new j. phys . * 15 * ( 2013 ) 065002 .
r. b. laughlin , `` quantized hall conductivity in two - dimensions , '' phys .
b * 23 * ( 1981 ) 5632 .
a. cappelli , g. v. dunne , c. a. trugenberger , g. r. zemba , `` conformal symmetry and universal properties of quantum hall states , '' nucl .
b * 398 * ( 1993 ) 531 .
s. ryu , j. e. moore , a. w. w. ludwig , `` electromagnetic and gravitational responses and anomalies in topological insulators and superconductors '' , phys .
b * 85 * ( 2012 ) 045104 . a. kitaev , `` periodic table for topological insulators and superconductor '' , adv .
landau memorial conference .
aip conference proceedings * 1134 * ( 2009 ) 22 ; a. p. schnyder , s. ryu , a. furusaki , a. w. w. ludwig , `` classification of topological insulators and superconductors in three spatial dimensions '' , phys . rev .
b * 78 * ( 2008 ) 195125 . x. -l .
qi , s. -c .
zhang , `` spin charge separation in the quantum spin hall state , '' phys .
* 101 * ( 2008 ) 086802 .
z. ringel , a. stern `` the @xmath398-anomaly and boundaries of topological insulators '' , preprint arxiv:1212.3796 .
e. witten , `` an @xmath399 anomaly '' phys .
b * 117 * ( 1982 ) 324 ; a. n. redlich , `` parity violation and gauge noninvariance of the effective gauge field action in three - dimensions '' phys .
d * 29 * ( 1984 ) 2366 ; s. deser , l. griguolo and d. seminara , `` effective qed actions : representations , gauge invariance , anomalies and mass expansions , '' phys .
d * 57 * ( 1998 ) 7444 .
j. polchinski , `` string theory .
1 : an introduction to the bosonic string , '' cambridge univ . press ( 1998 ) cambridge uk , vol i , chapter 8.4 .
t. gannon , `` u(1)-m modular invariants , n=2 minimal models , and the quantum hall effect , '' nucl . phys .
b * 491 * ( 1997 ) 659 .
hansson , c .- c .
chang , j.k .
jain , s. viefers , `` conformal field theory of composite fermions '' , phys .
* 98 * ( 2007 ) 076801 , `` composite - fermion wave functions as correlators in conformal field theory '' , phys .
b * 76 * ( 2007 ) 075347 ; a. cappelli , `` composite fermion wavefunctions derived by conformal field theory '' , j. phys .
a * 46 * ( 2013 ) 012001 .
a. cappelli , l. s. georgiev , i. t. todorov , `` parafermion hall states from coset projections of abelian conformal theories , '' nucl .
b * 599 * ( 2001 ) 499 .
a. cappelli , c. a. trugenberger , g. r. zemba , `` stable hierarchical quantum hall fluids as w(1+infinity ) minimal models , '' nucl .
b * 448 * ( 1995 ) 470 .
g. w. moore , n. read , `` nonabelions in the fractional quantum hall effect , '' nucl .
b * 360 * ( 1991 ) 362 .
n. read , e. rezayi , `` beyond paired quantum hall states : parafermions and incompressible states in the first excited landau level , '' phys .
b * 59 * ( 1999 ) 8084 .
o. m. sule , x. chen , s.ryu , `` symmetry - protected topological phases and orbifolds '' , preprint arxiv:1305.0700 . m. levin , z. c. gu `` braiding statistics approach to symmetry - protected topological phases '' , phys .
b * 86 * ( 2012 ) 115109 ; m. levin , `` protected edge modes without symmetry '' , preprint arxiv:1301.7355 ; z. c. gu , m. levin , `` the effect of interactions on 2d fermionic symmetry - protected topological phases with @xmath397 symmetry '' , preprint arxiv:1304.4569 .
m. a. metlitski , c. l. kane , m. p. a. fisher , a symmetry - respecting topologically - ordered surface phase of 3d electron topological insulators " , preprint arxiv:1306.3286 ; x. chen , l. fidkowski , a. vishwanath , symmetry enforced non - abelian topological order at the surface of a topological insulator " , preprint arxiv:1306.3250v2 ; p. bonderson , c. nayak , x .-
qi , `` a time - reversal invariant topological phase at the surface of a 3d topological insulator '' , preprint arxiv:1306.3230 . | the non - chiral edge excitations of quantum spin hall systems and topological insulators are described by means of their partition function .
the stability of topological phases protected by time - reversal symmetry is rediscussed in this context and put in relation with the existence of discrete anomalies and the lack of modular invariance of the partition function .
the @xmath0 characterization of stable topological insulators is extended to systems with interacting and non - abelian edge excitations .
.6 in partition functions and stability criteria of topological insulators 0.2 in andrea cappelli + _ infn + via g. sansone 1 , 50019 sesto fiorentino - firenze , italy _ enrico randellini + _ dipartimento di fisica + via g. sansone 1 , 50019 sesto fiorentino - firenze , italy _ .2 in |
the galactic center is the nuclear region of the nearest spiral galaxy , milky way .
the environment is unique in the galaxy because the region contains several peculiar objects .
first , sagittarius a@xmath0 ( sgr a@xmath0 ) is a counter part of the galactic center black hole ( gcbh ) in the regime from radio to x - ray , which is located very near the dynamical center of the galaxy ( e.g. ( * ? ? ?
* reid 2003 ) ) and has a mass of @xmath3m@xmath4 ( e.g. ( * ? ? ? * ghez 2008 ) ; ( * ? ? ?
* gillessen 2009 ) ) .
second , the circum - nuclear disk ( cnd ) " is a torus - like molecular gas rotating around sgr a@xmath0 , which is identified at the distance to a few pc ( e.g. ( * ? ? ?
* gsten 1987 ) ) .
a bundle of the ionized gas streams is located in the inner cavity of the cnd .
this is called galactic center mini - spiral ( gcms ) " ( e.g. ( * ? ? ?
* ekers 1983 ) ; ( * ? ? ?
* lo&claussen 1983 ) ) . the stretched appearance and kinematics with large velocity gradient suggest that it is a tentative structure surrounding sgr a@xmath0 . finally , the central cluster is a star cluster centered at sgr a@xmath0 .
though the cluster concentrates within @xmath5 pc , it contains @xmath6100 ob and wr stars ( e.g. ( * ? ? ?
* genzel 1996 ) ; ( * ? ? ?
* paumard 2006 ) ) .
there remains considerable controversy about the cluster origin .
it would be difficult to form the cluster in a way by which stars are usually formed in the galactic disk because of the following reasons .
the tidal force of sgr a@xmath0 must have a serious effect on the star formation because the minimum number h@xmath7 density for stabilization toward the tidally shearing is @xmath8h@xmath9 @xmath10 at @xmath5 pc ( ( * ? ? ?
* christopher 2005 ) , ( * ? ? ?
* montero - castao 2009 ) , ( * ? ? ?
* tsuboi 2011 ) ) .
in addition , the strong lyman continuum radiation from the early type stars in the cluster ionizes rapidly the ism in the region . because the star formation in the vicinity of the sgr a@xmath0 requires an additional mechanism to overcome the difficulties , it is still an open question how the central cluster is formed .
two distinct but compatible scenarios for the formation of the central cluster have been proposed .
one is current in - situ star formation in such extreme environment of the vicinity of sgr a@xmath0 .
previous observations have revealed that there are many molecular clumps in the cnd(e.g .
* montero - castao 2009 ) , ( * ? ? ?
* martn 2012 ) ) .
cloud - cloud collision in the cnd is proposed as the star formation mechanism in the region ( e.g. ( * ? ? ? * jalali 2014 ) ) .
unfortunately , we find no cradle dense molecular clouds around massive stars in the inner cavity of the cnd at present .
the other is that the molecular cloud is falling from the region considerably far from sgr a@xmath0 to the vicinity after star formation have already started in the cloud .
thus we searched the falling molecular cloud with star formation in the sgr a region .
emission line , * b * : in the sio @xmath11 emission line , * c * : in the h42@xmath2 recombination line .
the central velocity is @xmath12 km s@xmath13 .
the angular resolution is shown on the lower - left corner of the each panel as a filed oval .
the integration velocity widths are 10 km s@xmath13 in * a * and * b * and 20 km s@xmath13 in * c*. the contours show the continuum emission of the galactic center mini - spiral " at 100 ghz for comparison ( ( * ? ? ?
* tsuboi 2016a ) ) . ,
we have produced a 137 pointing mosaic of the 12-m array and a 52 pointing mosaic of the 7-m array ( aca ) which cover a @xmath14 area including the galactic center 50 km s@xmath13 molecular cloud ( 50mc ) " and the cnd in the cs @xmath1(@xmath15 ghz ) , and sio @xmath16(@xmath17 ghz ) emission lines by alma cy.1 ( 2012.1.00080.s , pi tsuboi , m. ) .
the frequency range also includes the h42@xmath2 recombination line(@xmath18 ghz ) .
we have detected these emission lines from both the two regions .
the data have an angular resolution of @xmath19 using natural weighting " in uv sampling .
the alma synthesized beam is approximately 4 times smaller than those of previous molecular line observations ( e.g. ( * ? ? ?
* montero - castao 2009 ) , ( * ? ? ?
* martn 2012 ) ) .
j0006 - 0623 , j1517 - 2422 , j717 - 3342 , j1733 - 1304 , j1743 - 3058 , j1744 - 3116 and j2148 + 0657 were used as phase calibrators .
the flux density scale was determined using titan , neptune and mars .
the calibration and imaging of the data were done by casa ( ( * ? ? ?
* mcmullin 2007 ) ) .
the continuum emission of the gcms and sgr a@xmath0 was subtracted from the spectral data using the casa task uvcontsub .
because the observation has a large time span of one year and seven months , the flux uncertainty is as large as 10 % .
figures 1a and 1b show the integrated intensity maps ( pseudo color ) of the cnd in the cs @xmath1 and sio @xmath11 emission lines , respectively .
they are cut from the large area mosaic data mentioned in the previous section .
the central velocity is @xmath12 km s@xmath13 both in the maps .
the velocity width of each panel is 10 kms@xmath13 .
the cs emission line is a dense molecular gas tracer with @xmath20(h@xmath7)@xmath21 cm@xmath13 , while the sio emission line is a tracer of shocked molecular gas with shock velocity of @xmath23 km s@xmath13 .
the angular resolutions are @xmath24 for the cs @xmath1 emission line and @xmath25 for the sio @xmath11 emission line , respectively .
they are shown on the lower - left corners of the two panels as filed ovals .
the contours in the figures show the continuum emission of the gcms at 100 ghz for comparison ( ( * ? ? ?
* tsuboi 2016a ) ) .
we found a filamentary molecular cloud extending from @xmath26 , @xmath27 to @xmath28 , @xmath29 in figure 1a .
the north end of the molecular cloud reaches beyond the cnd ; @xmath30 pc .
this molecular cloud is prominent in the velocity ranges from @xmath31 km s@xmath13 to @xmath32 km s@xmath13 . along
the velocity increases as approaching the south end .
although the filamentary molecular cloud is prominent in the cs 2 - 1 emission line , this had never been identified by existing telescopes because it is deeply embedded in the cnd . however , this is distinguishable clearly from other clouds belong to the cnd with the curved filamentary appearance revealed by the high angular resolution of alma . on the other hand ,
the filamentary molecular cloud is not prominent in the sio emission line except around the south end , @xmath33 , @xmath34 .
this indicates that the cloud suffers from strong shock around the south end although there is no evidence of strong shock in the other parts of the cloud .
the south end of the molecular cloud has the very wide velocity width of @xmath35 km s@xmath13 . while there is another component along the
western arc " of the gcms in figures 1a and 1b .
this component probably belongs to the cnd .
meanwhile , figure 1c shows the integrated intensity map ( pseudo color ) of the gcms in the h42@xmath2 recombination line ( pseudo color ) .
the h42@xmath2 recombination line is an ionized gas tracer .
the south end seems to be connected continuously to the north extension of the eastern arm " and northeastern arm " of the gcms ( ( * ? ? ?
* tsuboi 2016b ) ) .
although the velocity width of the north extension is as large as @xmath36 km s@xmath13 , the central velocity is @xmath37 km s@xmath13 ( ( * ? ? ?
* tsuboi 2016b ) ) and overlaps to that of the south end of the molecular cloud .
the central velocity of the molecular cloud is also continuous with that of the ionized gas in the arms .
the morphological and kinematic properties suggest that the molecular cloud continues to the ionized gas physically .
as mentioned above , the shocked gas traced by the sio emission is located around the boundary of these .
the molecular cloud is ionized by the lyman continuum from the central cluster as approaching to sgr a@xmath0 and expands at several hundreds times because the gas temperature increases from several @xmath38 k to @xmath39 k. therefore the interaction between the molecular gas and ionized gas probably produces a shock wave around the interface .
the filamentary molecular cloud is probably falling from the outer region to the sgr a region and being disrupted by the tidal shear of sgr a@xmath0 .
the star formation in the cloud may have already started by an external trigger like cloud - cloud collision when the cloud was in the region far from sgr a@xmath0 .
shock waves induced by cloud - cloud collision may form massive molecular cloud cores ( e.g. ) .
a possible remnant of such cloud - cloud collision is observed in the 50mc ( ( * ? ? ?
* tsuboi 2015 ) ) . in the case of nearly head - on collision
, the molecular clouds lose their angular momentums and begin to fall toward sgr a@xmath0 . moreover , the formed cores in the clouds can not be destroyed by the tidal force because of their high densities .
the massive cores quickly grow to ob stars .
the speculation is consistent with the fact that the ob stars are observed on the dust peaks in the gcms ( ( * ? ? ?
* tsuboi 2016a ) ) . while less massive cores gradually grows to low mass stars
. these may be still on the stage of protostars in the gcms .
recently , several half - shell - like ionized components are found near sgr a@xmath0 by jvla , which would be low mass protostars ( ( * ? ? ?
* yusef - zadeh 2015 ) ) .
they are located exclusively on the approaching tip of the northeastern arm " .
these suggest that the filamentary cloud and gcms play a role in transferring material including protostars from the outer region of the sgr a complex to the vicinity of sgr a@xmath0 .
some of the falling stars may be captured around sgr a@xmath0 and others may be scattered far from sgr a@xmath0 . | we found a molecular cloud connecting from the outer region to the galactic center mini - spiral ( gcms ) " which is a bundle of the ionized gas streams adjacent to sgr a@xmath0 .
the molecular cloud has a filamentary appearance which is prominent in the cs @xmath1 emission line and is continuously connected with the gcms .
the velocity of the molecular cloud is also continuously connected with that of the ionized gas in the gcms observed in the h42@xmath2 recombination line .
the morphological and kinematic relations suggest that the molecular cloud is falling from the outer region to the vicinity of sgr a@xmath0 , being disrupted by the tidal shear of sgr a@xmath0 and ionized by uv emission from the central cluster .
we also found the sio @xmath1 emission in the boundary area between the filamentary molecular cloud and the gcms .
there seems to exist shocked gas in the boundary area . |
inspired by the spectacular discoveries of oscillations of atmospheric @xcite , solar @xcite , and reactor neutrinos@xcite ( for recent reviews see @xcite ) the physics community worldwide is embarking on the next challenging problem , finding whether neutrinos are indeed majorana particles as many particle physics models suggest .
study of the neutrinoless double beta decay ( @xmath1 ) is the best potential source of information about the majorana nature of the neutrinos @xcite .
moreover , the rate of the @xmath1 decay , or limits on its value , can be used to constrain the neutrino mass pattern and the absolute neutrino mass scale , i.e. , information not available by the study of neutrino oscillations .
( the goals , and possible future directions of the field are described , e.g. , in the recent study @xcite .
the issues particularly relevant for the program of @xmath1 decay search are discussed in @xcite . ) the observation of @xmath1 decay would immediately tell us that neutrinos are massive majorana particles .
but without accurate calculations of the nuclear matrix elements it will be difficult to reach quantitative conclusions about the absolute neutrino masses and mass hierarchies and confidently plan new experiments . despite years of effort there is at present a lack of consensus among nuclear theorists how to correctly calculate the nuclear matrix elements , and how to estimate their uncertainty ( see e.g. @xcite ) .
since an overwhelming majority of published calculations is based on the quasiparticle random phase approximation ( qrpa ) and its modifications , it is worthwhile to try to see what causes the sizable spread of the calculated @xmath0 values .
does it reflect some fundamental uncertainty , or is it mostly related to different choices of various adjustable parameters ? if the latter is true ( and we believe it is ) can one find and justify an optimal choice that largely removes such unphysical dependence ?
in the previous paper @xcite we have shown that by adjusting the most important parameter , the strength of the isoscalar particle - particle force so that the known rate of the @xmath2-decay is correctly reproduced , the dependence of the calculated @xmath1 nuclear matrix elements @xmath0 on other things that are not a priori fixed , is essentially removed .
in particular , we have shown that this is so as far the number of single particle states included is concerned , and the choice of the different realistic representations of the nucleon @xmath4-matrix .
in @xcite we applied this procedure to the @xmath1 decay candidate nuclei , @xmath5ge , @xmath6mo , @xmath7te , and @xmath8xe . in the present work
we wish to expand and better justify the ideas presented in @xcite .
first , the method is systematically applied to calculate the nuclear matrix elements @xmath0 for most of the nuclei with known experimental @xmath2-decay rates .
second , the sensitivity of the results to variation of other model parameters is tested .
these are the axial vector quenching factor , commonly described as a modification of the constant @xmath3 , and the parameters that describe the effect of the short range correlations .
finally , arguments in favor of the chosen calculation method are presented and discussed .
provided that a virtual light majorana neutrino with the effective mass @xmath9 , @xmath10 is exchanged between the nucleons @xcite the half - life of the @xmath1 decay is given by @xmath11 is the precisely calculable phase - space factor , and @xmath12 is the corresponding nuclear matrix element .
thus , obviously , any uncertainty in @xmath12 makes the value of @xmath9 equally uncertain .
the elements of the mixing matrix @xmath13 and the mass - squared differences @xmath14 can be determined in oscillation experiments . if the existence of the @xmath1 decay is proved and the value of @xmath15 is found , combining the knowledge of the first row of the neutrino mixing matrix @xmath13 and the mass - squared differences @xmath14 , a relatively narrow range of absolute neutrino mass scale can be determined , independently of the majorana phases @xmath16 in most situations @xcite .
however , such an important insight is possible only if the nuclear matrix elements are accurately known .
the nuclear matrix element @xmath12 is defined as @xmath17 where @xmath18 are the wave functions of the ground states of the initial ( final ) nuclei .
we note that for @xmath19 nuclear matrix element @xmath12 coincides with @xmath20 of our previous work @xcite .
we have chosen this , somewhat awkward , parameterization so that we could later modify the value of @xmath3 and still use the same phase space factor @xmath21 that contains @xmath22 , tabulated e.g. in ref .
@xcite .
the explicit forms of the operators @xmath23 and @xmath24 are given in ref .
@xcite . in order to explain the notation used below we summarize here only the most relevant formulae .
we begin with the effective transition operator in the momentum representation @xmath25 here , @xmath26 and @xmath27 , h_t = g_a^2 \left[\frac{2}{3 } \frac{\vec{q}^2 } { \vec{q}^2 + m_{\pi}^2 } - \frac{1}{3 } \left ( \frac{\vec{q}^2}{\vec{q}^2 + m_{\pi}^2 } \right)^2 \right ] .
\label{eq : formf}\ ] ] for simplicity we do not explicitly indicate here the @xmath28 dependence of the form factor @xmath29 ( the usual dipole form is used for both @xmath30 and @xmath29 ) , and the terms containing @xmath31 .
the full expressions can be found in ref .
@xcite and are used in the numerical calculations .
note that the space part of the momentum transfer four - vector is used in eq .
( [ eq : formf ] ) since the time component is much smaller than @xmath32 .
the parts containing @xmath33 come from the induced pseudoscalar form factor @xmath34 for which the partially conserved axial - vector current hypothesis ( pcac ) has been used . in comparison with most of previous @xmath1 decay studies @xcite the higher order terms of the nucleon current ( in particular the induced pseudoscalar as well as the weak magnetism ) are included in the present calculation , resulting in a reduction of the nuclear matrix element by about 30% @xcite . in the numerical calculation of @xmath12 here
the summation over the virtual states in the intermediate nucleus is explicitly performed .
the relatively important role of the induced nucleon currents deserves a comment . in the charged current neutrino induced reactions the parts of the cross section containing @xmath34 are usually unimportant because they appear proportional to the outgoing charged lepton mass .
that comes about because the equation of motion for the outgoing on - mass - shell lepton is used . here
, this can not be done since the neutrino is virtual , and highly off - mass - shell . in order to calculate the matrix element in the coordinate space
, one has to evaluate first the ` neutrino potentials ' that in our case explicitly depend on the energy @xmath35 of the intermediate state , @xmath36 where @xmath37 stands for @xmath38 .
@xmath39 and @xmath40 , where @xmath41 are the spherical bessel functions .
the form factor combinations @xmath42 are defined in eq.([eq : formf ] ) , @xmath43 is the distance between the nucleons , @xmath44 is the nuclear radius , and @xmath45 are the ground state energies of the initial ( final ) nuclei .
we note that in ref .
@xcite the tensor potential is presented incorrectly with the spherical bessel function @xmath46 instead of @xmath47 .
however , the numerical results were obtained with the correct expression .
the individual parts of the matrix element @xmath12 , eq.([eq : m0nudef ] ) , are given by the expression : @xmath48_j || j^{\pi } k_f \rangle \langle j^{\pi } k_f | j^{\pi } k_i \rangle \langle j^{\pi } k_f || [ c_p^+ \tilde{c}_n]_j || 0_i^+ \rangle ~. \label{eq : long } \end{aligned}\ ] ] here we define only those symbols that are needed further , for full explanation see @xcite .
the summation over @xmath49 represents the summation over the different multipolarities in the virtual intermediate odd - odd nucleus ; such states are labeled by the indices @xmath50 when they are reached by the corresponding one - body operators acting on the initial or final nucleus , respectively .
the overlap factor @xmath51 accounts for the difference between them .
the operators @xmath52 contain the corresponding neutrino potentials and the relevant spin and isospin operators .
short range correlation of the two initial neutrons and two final protons are described by the jastrow - like function @xmath53 , @xmath54 where the usual choice is @xcite @xmath55 = 1.1 @xmath56 , @xmath57 = 0.68 @xmath56 .
( these two parameters are correlated . ) finally , the reduced matrix elements of the one - body operators @xmath58 ( @xmath59 denotes the time - reversed state ) depend on the bcs coefficients @xmath60 and on the qrpa vectors @xmath61 @xcite .
the difference between qrpa and rqrpa resides in the way these reduced matrix elements are calculated .
as in @xcite , the quasiparticle random phase approximation ( qrpa ) and its modification , the renormalized qrpa ( rqrpa ) , are used to describe the structure of the intermediate nuclear states virtually excited in the double beta decay .
we stress that in the qrpa and rqrpa one can include essentially unlimited set of single - particle states , labeled by @xmath62 , @xmath63 in eq.([eq : long ] ) , but only a limited subset of configurations ( iterations of the particle - hole , respectively two - quasiparticle configurations ) , in contrast to the nuclear shell model where the opposite is true . on the other hand , within the qrpa
there is no obvious procedure that determines how many single particle states one should include .
hence , various authors choose this crucial number ad hoc , basically for reasons of convenience .
as has been already shown in @xcite , a particular choice of the realistic residual two - body interaction potential has almost no impact on the finally calculated mean value and variance @xmath64 of @xmath12 , with the overwhelming contribution to @xmath64 coming from the choice of the single - particle basis size .
therefore , we perform the calculations here using @xmath4-matrix based only on the bonn - cd nucleon - nucleon potential . it is well known that the residual interaction is an effective interaction depending on the size of the single - particle ( s.p . ) basis .
hence , when the basis is changed , the interaction should be modified as well . for each nucleus in question three
single - particle bases are chosen as described in ref .
@xcite with the smallest set corresponding to @xmath65 particle - hole excitations , and the largest to about @xmath66 excitations .
energies are calculated with the coulomb corrected woods - saxon potential .
in qrpa and rqrpa there are three important global parameters renormalizing the bare residual interaction .
first , the pairing part of the interaction is multiplied by a factor @xmath67 whose magnitude is adjusted , for both protons and neutrons separately , such that the pairing gaps for the initial and final nuclei are correctly reproduced .
this is a standard procedure and it is well - known that within the bcs method the strength of the pairing interaction depends on the size of the s.p .
second , the particle - hole interaction block is renormalized by an overall strength parameter @xmath68 which is typically adjusted by requiring that the energy of the giant gt resonance is correctly reproduced .
we find that the calculated energy of the giant gt state is almost independent of the size of the s.p .
basis and is well reproduced with @xmath69 .
accordingly , we use @xmath70 throughout , without adjustment .
third , an important strength parameter @xmath71 renormalizes the particle - particle interaction ( the importance of the particle - particle interaction for the @xmath72 strength was recognized first in ref .
@xcite , and for the @xmath73 decay in @xcite ) .
the decay rate for both modes of @xmath73 decay is well known to depend sensitively on the value of @xmath71 .
this property has been used in @xcite to fix the value of @xmath71 for each of the s.p .
bases so that the known half - lives of the @xmath74 decay are correctly reproduced .
such an adjustment of @xmath71 , when applied to all multipoles @xmath49 , has been shown in @xcite to remove much of the sensitivity to the number of single - particle states , to the @xmath75 potential employed , and even to whether rqrpa or just simple qrpa methods are used .
this is in contrast to typical conclusion made in the recent past @xcite that the values of @xmath12 vary substantially depending on all of these things .
we believe that the @xmath76 decay rate is especially suitable for such an adjustment , in particular because it involves the same initial and final states as the @xmath77 decay .
moreover , the qrpa is a method designed to describe collective states as well as to obey various sum rules .
both double - beta decay amplitudes , @xmath1 and @xmath2 , receive contributions from many intermediate states and
using one of them for fixing parameters of qrpa seems preferable .
we will elaborate this point in the next section . while the above arguments are plausible , the issue of adequacy of the qrpa method in general , and the chosen way of adjusting the parameters , should be also tested using suitable simplified models that allow exact solution .
one such test was performed recently @xcite .
it involved two shells of varying separation , and a schematic interaction . from the point of view of the present work
, the model suggested that qrpa is an excellent approximation of the exact solution .
however , it turned out that the present method of parameter adjustment was not able to eliminate fully the effect of the higher , almost empty , shell on the @xmath1 decay matrix element .
it is not clear whether this is a consequence of the schematic nature of the model , or of some more fundamental cause . in any case
, the present results suggest that , within qrpa and rqrpa , the effects of the far away single particle states can be indeed eliminated , or at least substantially reduced .
it is well known that the calculated gamow - teller strength is larger than the experimental one . to account for this ,
it is customary to ` quench ' the calculated gt matrix elements .
formally , this could be conveniently accomplished by replacing the true value of the coupling constant @xmath3 = 1.25 by a quenched value @xmath78 1.0 .
it is not clear whether similar phenomenon exists for other multipoles , besides @xmath79 . to see the dependence on the chosen @xmath3 value , we use in this work both the unquenched and quenched value of the axial current coupling constant @xmath19 and @xmath80 , respectively ( for all multipoles ) .
the matrix elements @xmath12 calculated for the three s.p . bases and a fixed @xmath3 are relatively close to each other . as in @xcite , for each nucleus the corresponding average @xmath81 matrix elements ( averaged over the three choices of the s.p .
space ) is evaluated , as well as its variance @xmath64 . these quantities
( with the value of @xmath64 in parentheses ) are shown in table [ tab : t12 ] , columns 4 and 5 .
two lines for each nucleus represent the results obtained with @xmath19 ( the upper one ) and @xmath80 ( the lower one ) .
one can see that not only is the variance substantially less than the average value , but the results of qrpa , albeit slightly larger , are quite close to the rqrpa values .
furthermore , the ratio of the matrix elements calculated with different @xmath3 is closer to unity ( in most cases they differ only by @xmath8220% ) than the ratio of the respective @xmath3 squared ( 1.6 in our case ) .
the reason for such a partial compensation of the @xmath3-dependence is that the experimental @xmath83 for @xmath80 is larger than @xmath83 for @xmath19 after adjusting in both calculation the @xmath71 separately to the experimental @xmath2-decay transition probability .
correspondingly , one gets smaller adjusted value of @xmath71 leading to larger calculated @xmath0 .
thus , with the adopted choice of parameter fixing the resulting @xmath1 decay rate depends on the adopted @xmath3 markedly less than the naive scaling @xmath84 that would suggest a change by a factor of 2.44 .
naturally , the @xmath2 half - lives are known only with some uncertainty . to see how the experimental error in @xmath83 affects the calculated @xmath12 , the derivatives @xmath85 at the experimental value of @xmath83
are calculated .
finally , the errors induced by the experimental uncertainties in @xmath83 , @xmath86 , are given in column 6 of table [ tab : t12 ] .
another uncertainty is related to the treatment of the short range nucleon correlation .
the adopted form of the jastrow - type factor @xmath53 , eq .
( [ eq : short ] ) , is based on the work @xcite where a range of the values of the parameter @xmath55 is given .
we vary that parameter @xmath55 from 0.9 to 1.2 ( @xmath57 is fully determined by @xmath55 ) .
the calculated dependence of @xmath12 is shown in fig .
[ 0nbbfig:1 ] for @xmath5ge and @xmath6mo .
one can see that over this range of @xmath55 values the calculated results differ by only about 10% from the ones corresponding to @xmath87 used in the standard calculations . since @xmath83 is not affected by the short range repulsion , fig .
[ 0nbbfig:1 ] simply shows that the @xmath1 matrix elements @xmath12 are not very sensitive to reasonable variations of the parameter @xmath55 @xcite . combining the average @xmath81 with the phase - space factors ,
the expected half - lives ( for rqrpa and @xmath88 = 50 mev , the scale of neutrino masses suggested by oscillation experiments ) are also shown in table [ tab : t12 ] ( column 7 ) .
the information collected in table [ tab : t12 ] is presented in graphical form in fig .
[ 0nbbfig:2 ] . there
the averaged nuclear matrix elements for both methods and both choices of @xmath3 are shown along with their full uncertainties ( theoretical plus experimental ) .
the entries for @xmath7te are slightly different from the corresponding results in ref.@xcite .
this is so because in the present work we use the @xmath2 lifetime tentatively determined in the recent experiment @xcite , while in @xcite we used the somewhat longer @xmath2 lifetime based on the geochemical determination @xcite .
one can qualitatively understand why our chosen procedure stabilizes the @xmath12 matrix elements as follows : the @xmath83 matrix elements involve only the @xmath89 ( virtual ) states in the intermediate odd - odd nucleus . the nuclear interaction is such that the gamow - teller correlations ( spin one , isospin zero pairs ; after all the deuteron is bound and the di - neutron is not ) are very near the corresponding phase - transition in the @xmath89 channel ( corresponding to the collapse of the qrpa equations of motion ) .
the contributions of the @xmath89 multipole for both modes of the @xmath73 decay ( @xmath76 and @xmath77 ) depends therefore very sensitively on the strength of the particle - particle force , parameterized by @xmath71 . on the other hand , the @xmath12 matrix element , due to the presence of the neutrino propagator , depends on states of many multipolarities in the virtual intermediate odd - odd nucleus
the other multipoles , other than @xmath89 , correspond to small amplitudes of the collective motion ; there is no instability for realistic values of @xmath90 .
hence , they are much less sensitive to the value of @xmath71 . in the four panels of fig .
[ 0nbbfig:3 ] we show the @xmath71 dependence of the contribution of the @xmath89 multipole on one hand and of all the other multipoles added on the other hand .
remembering that the nominal value of @xmath71 is @xmath91 , one can clearly see the large difference in the corresponding slopes . by making sure that the contribution of the @xmath89 multipole is fixed
, we therefore stabilize the @xmath12 value . the fact that rqrpa essentially removes
the instability becomes then almost irrelevant thanks to the chosen adjustment of @xmath71 .
the effect of stabilization against the variation of the basis size is seen clearly if one plots directly @xmath12 versus @xmath83 , fig .
[ 0nbbfig:4 ] .
also the obtained multipole decompositions of @xmath12 plotted in fig .
[ 0nbbfig:5 ] show the essential stability of the partial contributions against variation of the basis size .
( note that in the case of @xmath92 our approach gives rather small matrix element and its uncertainty is even compatible with zero value .
the smallness of that matrix element is due to a large negative contribution of the @xmath89 multipole , see fig .
[ 0nbbfig:5 ] ) .
since the @xmath76 lifetime depends on the square of @xmath83 , there is an ambiguity in choosing the sign of the @xmath83 , and hence the corresponding value of @xmath71 . in this work , and in ref.@xcite ,
always the solution corresponding to the smaller @xmath71 value is chosen ( positive @xmath83 with the present phase convention ) .
it is worthwhile to justify such a choice .
there are several reasons why the smaller @xmath71 should be used .
first , qrpa and rqrpa are methods designed to describe small amplitude excitations around the mean field minimum .
were we to choose the larger value of @xmath71 , close or past the critical ` collapse ' value , the method would be less likely to adequately describe the corresponding states .
second , as we will show later , by choosing the larger @xmath71 the disagreement between the experimental and calculated rate of the single beta transitions from the lowest @xmath89 state in the intermediate nucleus would be far worse than with the smaller @xmath71 .
moreover , as shown in ref .
@xcite , only with the smaller @xmath71 can one successfully describe the systematic of single beta decay in a variety of nuclei .
it follows from the study of ref . @xcite that choosing the larger value of @xmath71 ( i.e. , the negative sign of @xmath83 ) would lead to a complete disagreement with the systematics of single
beta decays .
finally , there is also a pragmatic argument for such a choice . only with it ,
the @xmath77 nuclear matrix element becomes independent of the size of the single particle basis .
thus , this choice , admittedly ad hoc , removes the dependence on many more essentially arbitrary choices .
one can see in fig.[0nbbfig:5 ] that all multipoles @xmath49 ( see eq.([eq : long ] ) ) , with the exception of the @xmath89 and various very small entries , contribute with the same sign .
this suggests that uncertainties in one or few of them will have relatively minor effect .
it is instructive to see separately the effect of the sometimes neglected short - range repulsive nucleon - nucleon repulsion and of the induced weak nucleon currents .
these effects are shown in fig .
[ 0nbbfig:6 ] .
one can see in that figure that the conclusion of the relative role of different multipoles is affected by these terms .
for example , in @xmath6mo the @xmath93 multipole is the strongest one when all effects are included , while the @xmath94 becomes dominant when they are neglected .
it is worthwhile to point out that one can display various contributions to @xmath12 in an alternative , perhaps more revealing , way . instead of @xmath49 that represents virtual states in the intermediate odd - odd nucleus
, one can decompose the result in terms of @xmath95 , the angular momentum of the neutron - neutron ( and simultaneously proton - proton ) pair that undergoes the transition . in that case @xmath96
represents the ` pairing only ' contribution , while @xmath97 come from the ground state correlations , i.e. , contributions from higher seniority ( broken pairs ) states . in fig .
[ 0nbbfig:7 ] we show that these higher seniority states contributions consistently have the tendency to cancel the ` pairing only ' piece .
thus , the final @xmath12 is substantially less than the @xmath96 part only , signifying the importance of describing the ground state contributions properly .
( this tendency is a well known effect , for both modes of the @xmath73 decay .
it has been discussed , e.g. in ref .
one can not expect that the qrpa - like and shell model calculations will lead to identical results due to substantial differences between both approaches .
( however , we point out below that the results of the present approach and the shell model results differ relatively little whenever a comparison is possible . ) our goal in this section is to show that a majority of differences among various qrpa - like calculations can be understood .
in addition , we discuss the progress in the field and possible convergence of the qrpa results . based on our analysis
, we suggest that it is not appropriate to treat all calculated @xmath1-decay matrix elements at the same level , as it is commonly done ( see e.g. , @xcite ) , and to estimate their uncertainty based on their spread . here
we shall discuss the differences among different published qrpa and rqrpa results .
we shall not consider the problem of the proton - neutron pairing in the double beta decay @xcite and the self - consistent rqrpa ( srqrpa ) calculations @xcite .
a systematic study of the @xmath1-decay matrix elements within the srqrpa will be discussed in a forthcoming publication , where we show that within the considered approach of fixing nuclear structure input @xcite the results agree well with those obtained within the qrpa and the rqrpa .
representative examples of the nuclear matrix elements calculated by different authors within the qrpa and rqrpa for nuclei of experimental interest are collected in table [ tab : comp ] . in order to understand the differences between the entries ,
let us enumerate the main reasons leading to a spread of the published qrpa and rqrpa results : = = -=-5pt \i ) _
the quasiparticle mean field_. since the model space considered is finite , the pairing interactions have been adjusted to fit empirical pairing gaps based on the nuclear binding energies .
this procedure is followed practically by all authors
. however , it fails for closed proton ( @xmath98sn ) and neutron ( @xmath8xe ) shells .
some authors @xcite modify the single particle energies ( mostly spherical woods - saxon energies ) in the vicinity of the fermi surfaces to reproduce the low - energy quasiparticle spectra of the neighboring odd - mass nuclei .
\ii ) _ many - body approximations_. the rqrpa goes beyond the qrpa by partially taking into account pauli principle violation in evaluation of the bifermion commutators . based on that , one might expect that the rqrpa is more accurate than the unrenormalized qrpa .
calculations within solvable models @xcite support this conjecture .
\iii ) _ nucleon - nucleon interaction_. in the @xmath1-decay calculations both schematic zero - range @xcite and realistic interactions were considered . in ref .
@xcite @xmath4-matrix of the paris potential approximated by a sum of yukawa terms was used . the interaction employed by
the tuebingen group has been the brueckner @xmath4 matrix that is a solution of the bethe - goldstone equation with bonn ( bonn cd , argonne , nijmegen ) one boson exchange potential .
the results do not depend significantly on the choice of the nn interaction @xcite .
\iv ) _ the renormalization of the particle - hole interaction_. this is achieved by scaling the particle - hole part of the ( r)qrpa matrix by the parameter @xmath68 . that parameter is typically adjusted by requiring that the energy of some chosen collective state , often the giant gamow - teller ( gt ) resonance , is correctly reproduced . in ref .
@xcite it was found that the gt state is almost independent of the size of the model space and is well reproduced with @xmath99 . in refs .
@xcite @xmath100 was chosen and in ref .
@xcite @xmath101 was considered for @xmath102 nuclei .
the sensitivity of results to the change of @xmath68 were studied in ref .
@xcite where it was shown that for @xmath103 the change of the rqrpa @xmath1-decay matrix element does not exceed 10% .
\v ) _ the renormalization of the particle - particle interaction_. this is achieved by scaling the particle - particle part of the ( r)qrpa matrix by the parameter @xmath71 .
when the early qrpa calculation were performed , only a limited information about the experimental @xmath2-decay half - lives was available .
thus , in ref .
@xcite a probable @xmath71 window was estimated from @xmath104 decays of semi - magic nuclei . in another work @xcite systematics of the @xmath105-force
were investigated by an analysis of the single @xmath106 decays .
alternatively , in many works @xmath107 was chosen @xcite .
nowadays , the @xmath2-decay has been observed in ten nuclides , including decays into two excited states @xcite .
two recent papers used the @xmath2-decay half - lives to fix the strength of the particle - particle interaction of the nuclear hamiltonian . in ref .
@xcite it was used only for the @xmath108 channel leaving the particle - particle strength unrenormalized ( i.e. , @xmath109 ) in other channels . in our previous paper @xcite , and in the current work , @xmath71 is adjusted so that the @xmath2-decay rate is correctly reproduced .
the same @xmath71 is used for all multipole channels of the particle - particle interaction . since information that can be used to adjust the @xmath71 value exists only for the @xmath89 channel
, we can not make a separate adjustment for other multipoles .
our economical choice is then to use the same @xmath71 for all multipoles .
that makes the hamiltonian as simple as possible , and preserves the relative strength of different multipoles of the realistic starting point interaction .
on the other hand , some authors prefer to fix the @xmath71 value to the @xmath110 decay transition of the ground state of the intermediate nucleus @xcite .
such procedure can be carried out , however only for three nuclear systems ( a=100 , 116 and 128 ) where @xmath89 is the ground state of the intermediate nucleus . since forbidden decays
are less well understood than the allowed ones it is difficult to rely on them in the case of other nuclear systems .
( we discuss in more detail the differences between these two approaches of fixing the parameter @xmath71 in the next section . )
\vi ) _ the size of the model space .
_ a small model space comprising usually of two major shells was often used in the calculation of the nuclear matrix elements @xcite .
significantly larger model space has been used in refs.@xcite consisting of five major shells . as a rule
the results obtained for the large model space are reduced when compared to those for the small model space for the same value of @xmath111 .
the dependence of the calculated nuclear matrix elements on the size of the model space has been studied in ref.@xcite and in @xcite .
rather different conclusions using the same procedure of fixing @xmath71 was reached .
the origin of such discrepancy is not understood . in ref .
@xcite relatively small differences between model spaces were considered and a large effect on the @xmath1-decay matrix element was found . in particular , by adding the @xmath112 shell to the @xmath113 oscillator shell model space the @xmath1-decay matrix element of @xmath114 decreased by a factor of about 3 ( see table [ tab : comp ] ) .
thus , in ref .
@xcite the single - particle levels lying far from the fermi surface seem to influence strongly the decay rate .
in contrast , in ref .
@xcite and in the present work results essentially independent on the size of the basis have been found even with significantly different sizes of the model space .
\vii ) _ the closure approximation .
_ the @xmath1-decay matrix elements were usually calculated using the closure approximation for intermediate nuclear states @xcite . within this
approximation energies of intermediate states ( @xmath115 ) are replaced by an average value ( @xmath116 ) , and the sum over intermediate states is taken by closure , @xmath117 .
this simplifies the numerical calculation drastically .
the calculations with exact treatment of the energies of the intermediate nucleus were presented in ref .
the effect of the closure approximation was studied in details in ref .
it was found that the differences in nuclear matrix elements are within @xmath118 .
this is so because the virtual neutrino has an average momentum of @xmath119 mev , considerably larger than the differences in nuclear excitation energies .
\viii ) _ the axial - vector coupling constant @xmath3 . _ the axial - vector coupling constant or in other words , the treatment of quenching , is also a source of differences in the calculated nuclear matrix elements .
the commonly adopted values are @xmath80 @xcite and @xmath19 @xcite . however , as shown in table [ tab : comp ] , if @xmath71 is fixed to the @xmath2-decay half - life , the effect of @xmath3 modification is smaller , of order @xmath118 .
\ix ) _ the two - nucleon short - range correlations ( s.r.c.)_. in majority of calculations the short - range correlations between two nucleons are taken into account by multiplying the two - particle wave functions by the correlation function @xcite : @xmath120 with @xmath121 @xmath122 and @xmath123 @xmath122 .
the sensitivity of results to the change of this parameters have been discussed in the previous section .
it is also known that the @xmath1-decay matrix element is more affected ( reduced ) by the s.r.c . at higher @xmath71 values .
we note that the s.r.c . were not taken into account within the approach which was developed in ref.@xcite and used in the recent publication @xcite . for the realistic @xmath71 values neglecting s.r.c . would lead to an increase in @xmath0 by factor of about 2 .
\x ) _ the higher order terms of the nucleon current . _
the momentum dependent higher order terms of nucleon current , namely the induced pseudoscalar and weak magnetism , were first considered in connection with the light neutrino mass mechanism of the @xmath1-decay in @xcite .
their importance is due to the virtual character of the exchanged neutrino , with a large average momentum of @xmath82 100 mev .
the corresponding transition operators have different radial dependence than the traditional ones .
it is worth mentioning that with modification of the nucleon current one gets a new contribution to neutrino mass mechanism , namely the tensor contribution .
the corrections due to the induced pseudoscalar nucleon current vary slightly from nucleus to nucleus , and in all cases represent about @xmath124 reduction . in the present calculation
the effect of higher order terms of nucleon current is taken into account .
\xi ) _ the finite size of the nucleon _ is taken into account via momentum dependence of the nucleon form - factors .
usually this effect on the @xmath1-decay matrix elements associated with light neutrino exchange is neglected @xcite , since it is expected to be small . in the presented calculation
it is taken account by assuming phenomenological cutoff ( see @xcite and references therein ) .
we found that by considering the finite nucleon size the value of @xmath12 is reduced by about 10% . in refs .
@xcite the nucleon current is treated in a different way than in other double beta decay studies , namely it is evaluated from the quark level using relativistic quark wave functions .
however , the momentum dependence of corresponding nucleon form - factors is not shown , so a comparison is difficult . comparing the present results with those of ref .
@xcite for the same nuclear structure input suggests that the agreement might be achieved only if a very low cutoff is introduced .
\xii ) _ the overlap factor of intermediate nuclear states_. this factor is introduced since the two sets of intermediate nuclear states generated from initial and final ground states are not identical within the considered approximation scheme .
a majority of calculations uses a simple overlap factor @xmath125 , inspired by the orthogonality condition of rpa states generated from the same nucleus .
however , the double beta decay is a two - vacua problem .
the derivation of the overlap matrix within the quasiboson approximation scheme was performed in refs.@xcite .
it was shown that the overlap factor of the initial and final bcs vacua is an integral part of the overlap factor of the intermediate nuclear states @xcite .
this bcs overlap factor , about 0.8 for spherical nuclei , is commonly neglected in the calculation of nuclear matrix elements .
if one fits @xmath71 to the experimental @xmath2-transition probability , the effect of the bcs overlap factor is significantly reduced .
hence it is neglected in the present work .
this factor has been found to be important in the calculation within the deformed qrpa .
\xiii ) _ the nuclear shape .
_ until now , in all qrpa - like calculations of the @xmath1-decay matrix elements the spherical symmetry was assumed as the majority of nuclei of experimental interest are nearly spherical .
the effect of deformation on the @xmath1-decay matrix elements has not been studied as of now .
recently , the @xmath2-decay matrix elements were calculated within the deformed qrpa with schematic forces @xcite . it was found that differences in deformation between initial and final nuclei have a large effect on the @xmath2-decay half - life .
one could expect that a similar mechanism of suppression of nuclear matrix elements is present also in the case of the @xmath1-decay .
it goes without saying that a further progress concerning this topics is highly desirable .
the deformed qrpa might be the method of choice for the description of double beta decay of heavy nuclei like @xmath126nd , or @xmath127gd .
however , the deformation of many medium heavy nuclear systems are also noticeable ( see table ii of @xcite ) .
the main difficulty of the deformed rqrpa calculation is the fact that by going from the spherical to the deformed nuclear shapes the number of configurations increases drastically . the present table [ tab : comp ] ,
is organized as follows : i ) the results with and without inclusion of higher order terms of nucleon current are separated .
we point out that in refs .
@xcite the contribution from the weak - magnetism was taken into account .
it is , however , negligible and we present their results without this contribution .
ii ) we indicate those matrix elements , which differ significantly from the present ones or other previous calculations ( denoted as sk-01 @xcite and cs-03 @xcite ) .
we shall discuss this problem in detail below .
iii ) within the sub - blocks the nuclear matrix elements are presented in ascending order following the year they appeared .
many nuclear matrix elements in table [ tab : comp ] were calculated from the published ratios of fermi @xmath128 and gamow - teller @xmath129 contributions and the absolute values of @xmath130 presented in units of @xmath131 , i.e. , not scaled with nuclear radius @xcite we assumed that @xmath132 with @xmath133 fm .
however , ref .
@xcite used @xmath134 fm , so we rescaled the results in order to compare them better with ours .
the differences between many calculations are understandable just from the way @xmath71 was fixed , the considered size of the model space , the inclusion of the short - range correlation , the way the finite nucleon size is taken into account and other minor effects .
some nuclei , like @xmath6mo , exhibit more sensitivity to these effects than others , such as @xmath5ge or @xmath135se .
that is confirmed by our numerical studies .
the calculations of the @xmath1-decay matrix elements by civitarese and suhonen @xcite ( denoted as cs-03 in table [ tab : comp ] ) deserves more comments .
the authors performed them within the approach suggested in ref .
@xcite , ( skf-91 ) which employs the nucleon current derived from the quark wave functions . in this approach
the two nucleon short - range - correlations are not taken into account .
but , the contribution from higher order terms of nucleon current is studied .
unlike the present results the authors of ref.@xcite claim that contribution from the induced pseudoscalar coupling is a minor one and they do not include it . at the same time , they find that the contribution from the weak - magnetism is negligible @xcite , in agreement with our result and with the @xmath31 scaling . in ref .
@xcite the @xmath71 was taken to be unity .
the extension of this work for the case when @xmath71 is adjusted to reproduce the single @xmath72-decay decay amplitudes was presented in ref.@xcite ( as-98 ) where also the effect of adjustment of the single particle energies to agree better with the spectroscopic data of odd - mass nuclei ( @xmath136 ) was studied .
since the adopted value of @xmath3 used for fixing @xmath71 was not given , in table [ tab : comp ] we present the corresponding @xmath12 for both @xmath137 and @xmath138 .
( from the related article @xcite it seems @xmath137 was considered . ) in ref .
@xcite ( cs-03 ) the nuclear matrix elements are calculated in the same way as in @xcite , however , the obtained results differ significantly ( see table [ tab : comp ] ) from each other . for some nuclei
the difference is as large as a factor of two .
there is no discussion of this there or in the later refs .
it is worth noticing that the largest matrix element in @xcite is found for the @xmath1-decay of @xmath8xe .
this disagrees with the results of other authors ( see table [ tab : comp ] ) .
the reduction of the @xmath1-decay of the @xmath8xe is explained by the closed neutron shell for this nucleus .
a sharper fermi surface leads to a reduction of this transition .
altogether , the matrix elements of refs.@xcite are noticeably larger than the present ones .
most of that difference can be attributed to the neglect of the short range nucleon - nucleon repulsion and of the higher order effects in the nucleon weak current in these papers . as the above list of 13 points shows
, there are many reasons , some more important than others , that might cause a difference between various calculated @xmath1 matrix elements . clearly
, when some authors do not include effects that should be included ( e.g. the short range correlations or the higher order terms in the nucleon current ) their results should be either corrected or convincing arguments should be given why the chosen procedure was adopted .
other effects on the list are correlated , like the size of the model space and the renormalization of the particle - particle interaction . again
, if those correlations are not taken into account , erroneous conclusion might be drawn .
yet other effects are open to debate , like the quenching of @xmath3 or the adopted method of adjusting @xmath71 . in our previous work @xcite , and in the present one ,
we show that our chosen way of renormalization removes , or at least greatly reduces , the dependence of the final result on most of the effects enumerated above .
our choice of @xmath71 adjustment , to the experimental @xmath2 rate was used also in refs.@xcite .
let us comment on their results .
note , that the higher order corrections ( h.o.c . ) to the weak nucleon current were neglected in those works , hence one expects a @xmath82 30% discrepancy right away . in ref .
@xcite only the @xmath73 decay of @xmath5ge was considered within both the qrpa and rqrpa .
only one s.p .
basis was used , corresponding to the small one in our notation .
the result was @xmath139 within qrpa ( rqrpa ) . if one takes into account that h.o.c .
as a rule reduce @xmath0 by about 30% and the fact that @xmath82 10% larger nuclear radius was used , one concludes that these matrix elements should be multiplied by @xmath82 0.6 in order to compare them with our calculations .
this results in @xmath140 which are in a very good agreement with our calculated @xmath141 for this case . hence qrpa and
rqrpa gave quite similar results in that case .
in contrast , the conclusions in ref .
@xcite are quite different from ours ; the authors found significant dependence of their results on both the nuclear model and on the s.p .
basis size , in spite of the adjustment of @xmath71 .
the reason for these differences is unknown . in an attempt to understand the origin of them we comment here on features that , in our judgment , require further discussion .
\1 ) we have already pointed out that the @xmath0 for @xmath5ge in ref .
@xcite changes with changing the single - particle basis within qrpa - from 4.45 ( 9 levels ) to 1.71 ( 12 levels ) , or more than 2.5 times .
this was obtained by adding just the deep - lying @xmath142 shell , a surprising result .
\2 ) the last of the three papers contains most information and we can use their figs . 1 - 14 with @xmath83 and figs .
15 - 25 with @xmath0 , in order to compare the numbers in the first ( qrpa ) and the last ( sqrpa ) columns of their table 3 .
we were able to reproduce the entries in the qrpa column , but almost all entries in the sqrpa column , apart from those for @xmath135se and @xmath114 , seem inconsistent .
the most striking difference is for @xmath8xe . by inspecting fig . 7 in @xcite
, one finds that the appropriate @xmath71 for sqrpa is about 1.08 ( lines a and b ) .
now , by going to fig .
24 one can see that the corresponding values of @xmath0 are about 2.5 ( lines c and d ) while the entries in table 3 are 0.98 and 1.03 .
\3 ) some of the results in @xcite ( see fig.1 , lines a and b , sqrpa ; fig.2 , lines a and b , sqrpa , fig.5 , lines c and d , qrpa ) , fig.7 , lines c and d , qrpa ) contradict the generally accepted conclusion that for all rpa - like approaches the functions @xmath143 calculated with the larger basis cross zero faster than the ones obtained with the smaller one .
thus , given these apparent inconsistencies , it is difficult to draw any conclusions from the comparison of our and ref .
@xcite results . as pointed out above
, one can not expect a perfect agreement between the present result and the large scale shell model results @xcite .
the approximations are different , and the shell model calculations do not include several multipoles ( which typically enhance the matrix elements , see fig .
[ 0nbbfig:5 ] ) .
yet , remarkably , the present results and the published large scale shell model results agree with each other considerably better than the various entries in table [ tab : comp ] .
so , using the published values , @xcite , we find @xmath12 values 1.5 , 2.1 , 1.1 , and 0.7 for @xmath5ge , @xmath135se , @xmath7te and @xmath8xe , while rqrpa gives 2.4 , 2.1 , 1.5 , and 0.7 - 1.0 for the same nuclei .
however , it appears that the more recent shell model results @xcite for @xmath7te and @xmath8xe give larger values while at the same time overestimating to some degree the rate of the @xmath2 decay . in any case , we find this comparison encouraging .
ideally , the chosen nuclear structure method should describe all , or at least very many , experimental data and do that without adjustments . as described above that
is not the case of qrpa or rqrpa .
the interaction used is an effective interaction , and various parameters ( @xmath144 ) are adjusted . in particular , the parameters @xmath67 and @xmath71 are adjusted on the case - by - case basis in the present approach .
even then the method is not able to describe well all relevant weak transitions .
in particular , it is sometimes impossible to describe simultaneously the @xmath2 decay rate as well as the @xmath110 and @xmath106 matrix elements connecting the @xmath89 ground states of the intermediate nucleus with the ground states of the final and initial nuclei ( @xmath6mo is a well known example of this problem , see e.g. @xcite ) . empirically , the transitions through the @xmath89 ground state of @xmath6tc , @xmath98 in , and @xmath145i seemingly account for most of the @xmath2 matrix element ( this is the so called ` single state dominance ' @xcite ) .
thus it appears that these single @xmath72 transitions are particularly relevant .
based on such considerations , suhonen @xcite suggested that the @xmath110 matrix element is more suitable source for the @xmath71 adjustment than the @xmath2 decay .
below we explain why we prefer the chosen method of parameter adjustment . on the other hand , we also argue that the adjustment based on the @xmath110 transitions might give results not far from ours , provided the independence on other adjustments can be proved .
first , it is not really true that the first @xmath89 state is the only one responsible for the @xmath2 decay .
this is illustrated for the cases of @xmath5ge and @xmath6mo in fig .
[ 0nbbfig:8 ] .
even though for @xmath6mo the first state contributes substantially , higher lying states give non - negligible contribution . and in @xmath5ge many @xmath89 states give comparable contribution . thus , to give preference to the lowest state is not well justified , the sum is actually what matters . at the same time , the dilemma that the @xmath110 and @xmath106 matrix elements move with @xmath71 in opposite directions makes it difficult to choose one of them .
it seems better to use the sum of the products of the amplitudes , i.e. the @xmath2 decay . at the same time , the contribution of the @xmath89 multipole to the @xmath1 matrix element and the corresponding @xmath2 matrix element are correlated , even though they are not identical , as shown in fig .
[ 0nbbfig:8 ] . making sure that the @xmath2 matrix element agrees with its experimental value constrains the @xmath89 part of the @xmath1 matrix element as well . in fig .
[ 0nbbfig:9 ] we show the running sum contributions to the @xmath1 matrix elements in @xmath5ge and @xmath6mo , separated into multipoles , and for the total . such a sum , even for the @xmath89 component , is rather different that the similar staircase for the @xmath2 ; there is no single state dominance .
thus , again , it is not obvious that it is best to choose any one particular state or transition for the adjustment .
moreover , while we have demonstrated that adjusting @xmath71 to the @xmath2 rate removes the dependence on other parameters , a similar proof was not given in @xcite . finally , as seen in fig . [ 0nbbfig:3 ] the slope of the @xmath89 multipole component of the @xmath1 matrix element far exceeds the slope of the other multipoles .
hence , making sure that the whole @xmath89 multipole contribution is correct is crucial .
thus , we prefer the method of adjustment used in the present work .
that is so not only for the reasons shown above , but also because it is much more general as stressed already , while the ground state of the odd - odd nucleus has @xmath89 only in @xmath6tc , @xmath98 in , and @xmath145i .
however , as demonstrated in the last figure , fig .
[ 0nbbfig:10 ] , the adjustment proposed in ref .
@xcite , namely to choose the parameter @xmath71 based on the @xmath110 decay of the intermediate @xmath89 state , would not be drastically different compared with the procedure used in the present work .
as one could see , the resulting @xmath71 are similar ( but not identical ) and the dependence on the s.p .
basis will be also reduced ( that important feature was not demonstrated in ref .
@xcite , unfortunately ) . however , as the upper panel shows , it is difficult to describe the other component of that decay , the @xmath106 amplitude .
that is an obvious drawback of the qrpa method ; it is never meant to describe in detail properties of non - collective states .
but that is less relevant for the description of integral quantities that depend on sums over many states . as pointed out earlier
, we always choose the @xmath71 corresponding to the positive @xmath83 value .
in the upper panel of fig . [ 0nbbfig:10 ]
one can clearly see that were we to choose the other possibility , i.e. , the @xmath71 corresponding to the negative @xmath83 , the disagreement with the single beta decay would be considerably worse .
it is important also to notice that if pauli principle violation is restored ( e.g. within the rqrpa ) one finds that the solution corresponding to the negative value of @xmath83 is out of physical interval of @xmath71 .
this has been confirmed also in a schematic model by presenting the solution of the qrpa with full inclusion of the pauli exclusion principle @xcite .
as pointed out above , it was also found recently @xcite that for negative value of @xmath83 the correspondence with particle - particle strength from systematic studies of the single beta decay @xcite is not achieved . in this section
we have summarized the arguments why we believe that the procedure of adjustment used in the present work is preferable to the procedure advocated in ref.@xcite . at the same time , we suggest that there is no fundamental difference between the two ; both are used to fix the fast varying contribution of the @xmath89 multipole .
the bulk of the matrix element is associated with the other multipoles , and their effect is much less dependent on relatively small variations of the parameter @xmath71 . as we stressed already above
, a substantial part of the differences in the calculated values of the @xmath1 decay matrix elements has its origin not in the choice of @xmath71 or other parameters , but in the neglect of the short range nucleon - nucleon repulsion and of the induced weak currents in some papers , and their inclusion in other papers ( like the present one ) . until a consensus on the treatment of these physics issues
is reached , the differences in the calculated matrix elements can not be avoided .
we have shown that the procedure suggested in our previous work , ref .
@xcite , is applicable to essentially all nuclei with known @xmath2 decay lifetimes . adjusting the strength of the particle - particle neutron - proton force @xmath71 in such a way that the experimental @xmath2 decay rate is correctly reproduced removes much of the dependence on the size of the single - particle basis and whether qrpa or rqrpa is used .
here we also show that the quenching of the axial current matrix elements , parameterized by the reduction of the coupling constant @xmath3 , also leaves the resulting @xmath1 matrix elements almost unchanged ; they become insensitive to the variations of parameters describing the short - range nucleon - nucleon correlations as well .
thus , the resulting @xmath1 matrix elements acquire well defined values , free of essentially arbitrary choices .
we also present arguments while we believe that the chosen procedure of adjusting the interaction is preferable to other proposed ways of adjustment .
we then summarize many published qrpa and rqrpa results and discuss their similarities and differences .
we show that in most , albeit not all , cases these differences can be understood .
we present an exhaustive list of reasons why individual calculated @xmath1 nuclear matrix elements , evaluated within qrpa or rqrpa , might differ from each other .
that list then can serve as a guide to readers of the past and future papers devoted to the subject .
comparison between the results of different qrpa / rqrpa calculations would be facilitated if authors of future publications specify in detail what choices of explicit and implicit adjustable parameters they made , and discuss the dependence of their result on their particular choice .
we believe that by following these suggestions a consensus among the practitioners of qrpa / rqrpa could be reached and most of the spread between the calculated nuclear matrix elements , that causes much confusion in the wider physics community , would be shown to be essentially irrelevant . to reach a convergence of the results obtained
using qrpa / rqrpa is clearly just an important step on the way to reliable and correct @xmath1 decay nuclear matrix elements .
exploring the structure of the intermediate odd - odd nuclei by the charge exchange reactions would create an opportunity to test the nuclear models more thoroughly .
further progress in nuclear shell model calculations and in the exploration of the exactly solvable models could also point the way towards the ultimate solution of this important problem .
discussions with jon engel are gratefully acknowledged . the work of f. . and v. r. was supported in part by the deutsche forschungsgemeinschaft ( grants 436 slk 17/298 , tu 7/134 - 1 and fa67/28 - 2 , respectively ) .
we thank also the eu ilias project under the contract rii3-ct-2004 - 506222 .
the work of p. v. was supported by the u.s .
doe , stanford university , slac and kipac .
f. imkovic , l. pacearescu , and a. faessler , nucl .
phys . a * 733 * , 321 ( 2004 ) ; r. alvarez - rodriguez , p. sarriguren , e. moya de guerra , l. pacearescu , a. faessler and f. imkovic , phys .
c * 70 * , 064309 ( 2004 ) .
a. garcia et al . , phys .
c * 47 * , 2910 ( 1993 ) ; h. ejiri and h. toki , j. phys . soc
. jpn . * 65 * , 7 ( 1996 ) ; m. bhattacharaya et al .
c * 58 * , 1247 ( 1998 ) ; o. civitarese and j. suhonen , nucl .
a * 653 * , 321 ( 1999 ) ; f. imkovic , p. domin , and s.v .
semenov , j. phys .
g * 27 * , 2233 ( 2001 ) ; p. domin , s.g .
kovalenko , f. imkovic , and s.v .
semenov , nucl .
a * 753 * , 337 ( 2005 )
. .averaged @xmath1 nuclear matrix elements @xmath81 and their variance @xmath64 ( in parentheses ) evaluated in the rqrpa and qrpa . in column 6 the variance @xmath146 of the @xmath1-decay matrix element due to uncertainties in the measured @xmath2-decay half - live @xmath147 is given .
@xmath148 and @xmath3 denote the @xmath2-decay nuclear matrix element deduced from @xmath147 and axial - vector coupling constant , respectively . in column 7
the @xmath1 half - lives evaluated with the rqrpa average nuclear matrix element and for assumed @xmath9 = 50 mev are shown . for @xmath8xe
there are four entries ; the upper two use the upper limit of the @xmath76 matrix element while the lower two use the ultimate limit , vanishing @xmath76 matrix element .
@xmath126nd is included for illustration .
it is treated as a spherical nucleus ; deformation will undoubtedly modify its @xmath77 matrix element . [
cols="<,^,^,^,^,^,^ " , ] | the nuclear matrix elements @xmath0 of the neutrinoless double beta decay ( @xmath1 ) of most nuclei with known @xmath2-decay rates are systematically evaluated using the quasiparticle random phase approximation ( qrpa ) and renormalized qrpa ( rqrpa ) .
the experimental @xmath2-decay rate is used to adjust the most relevant parameter , the strength of the particle - particle interaction .
new results confirm that with such procedure the @xmath0 values become essentially independent on the size of the single - particle basis .
furthermore , the matrix elements are shown to be also rather stable with respect to the possible quenching of the axial vector strength parametrized by reducing the coupling constant @xmath3 , as well as to the uncertainties of parameters describing the short range nucleon correlations .
theoretical arguments in favor of the adopted way of determining the interaction parameters are presented .
furthermore , a discussion of other implicit and explicit parameters , inherent to the qrpa method , is presented .
comparison is made of the ways these factors are chosen by different authors .
it is suggested that most of the spread among the published @xmath1 decay nuclear matrix elements can be ascribed to these choices . |
humans can effortlessly and rapidly recognize surrounding objects @xcite , despite the tremendous variations in the projection of each object on the retina @xcite caused by various transformations such as changes in object position , size , pose , illumination condition and background context @xcite .
this invariant recognition is presumably handled through hierarchical processing in the so - called ventral pathway .
such hierarchical processing starts in v1 layers , which extract simple features such as bars and edges in different orientations @xcite , continues in intermediate layers such as v2 and v4 , which are responsive to more complex features @xcite , and culminates in the inferior temporal cortex ( it ) , where the neurons are selective to object parts or whole objects @xcite . by moving from the lower layers to the higher layers , the feature complexity , receptive field size and transformation invariance increase , in such a way that the it neurons can invariantly represent the objects in a linearly separable manner @xcite .
another amazing feature of the primates visual system is its high processing speed . the first wave of image - driven neuronal responses in it appears around 100 ms after the stimulus onset @xcite .
recordings from monkey it cortex have demonstrated that the first spikes ( over a short time window of 12.5 ms ) , about 100 ms after the image presentation , carry accurate information about the nature of the visual stimulus @xcite .
hence , ultra - rapid object recognition is presumably performed in a feedforward manner @xcite . moreover , although there exist various intra- and inter - area feedback connections in the visual cortex , some neurophysiological @xcite and theoretical @xcite studies have also suggested that the feedforward information is usually sufficient for invariant object categorization . appealed by the impressive speed and performance of the primates visual system ,
computer vision scientists have long tried to `` copy '' it .
so far , it is mostly the architecture of the visual system that has been mimicked .
for instance , using hierarchical feedforward networks with restricted receptive fields , like in the brain , has been proven useful @xcite . in comparison , the way that biological visual systems learn the appropriate features has attracted much less attention .
all the above - mentioned approaches somehow use non biologically plausible learning rules .
yet the ability of the visual cortex to wire itself , mostly in an unsupervised manner , is remarkable @xcite . here
, we propose that adding bio - inspired learning to bio - inspired architectures could improve the models behavior . to this end , we focused on a particular form of synaptic plasticity known as spike timing - dependent plasticity ( stdp ) , which has been observed in the mamalian visual cortex @xcite .
briefly , stdp reinforces the connections with afferents that significantly contributed to make a neuron fire , while it depresses the others @xcite .
a recent psychophysical study provided some indirect evidence for this form of plasticity in the human visual cortex @xcite . in an earlier study @xcite , it is shown that a combination of a temporal coding scheme where in the entry layer of a spiking neural network the most strongly activated neurons fire first with stdp leads to a situation where neurons in higher visual areas will gradually become selective to complex visual features in an unsupervised manner .
these features are both salient and consistently present in the inputs .
furthermore , as learning progresses , the neurons responses rapidly accelerates .
these responses can then be fed to a classifier to do a categorization task . in this study , we show that such an approach strongly outperforms state - of - the - art computer vision algorithms on view - invariant object recognition benchmark tasks including 3d - object @xcite and eth-80 @xcite datasets .
these datasets contain natural and unsegmented images , where objects have large variations in scale , viewpoint , and tilt , which makes their recognition hard @xcite , and probably out of reach for most of the other bio - inspired models @xcite .
yet our algorithm generalizes surprisingly well , even when `` simple classifiers '' are used , because stdp naturally extracts features that are class specific .
this point was further confirmed using mutual information @xcite and representational dissimilarity matrix ( rdm ) @xcite .
moreover , the distribution of objects in the obtained feature space was analyzed using hierarchical clustering @xcite , and objects of the same category tended to cluster together .
the algorithm we used here is a scaled - up version of the one presented in @xcite .
essentially , many more c2 features and iterations were used .
our code is available upon request .
we used a five - layer hierarchical network @xmath0 @xmath1 @xmath2 @xmath1 @xmath3 @xmath1 @xmath4 @xmath1 @xmath5 , largely inspired by the hmax model @xcite ( see fig .
[ model_figure ] ) .
specifically , we alternated simple cells that gain selectivity through a sum operation , and complex cells that gain shift and scale invariance through a max operation .
however , our network uses spiking neurons and operates in the temporal domain : when presented with an image , the first layer s @xmath0 cells , detect oriented edges and the more strongly a cell is stimulated the earlier it fires .
these @xmath0 spikes are then propagated asynchronously through the feedforward network .
we only compute the first spike fired by each neuron ( if any ) , which leads to efficient implementations .
the justification for this is that later spikes are probably not used in ultra - rapid visual categorization tasks in primates @xcite .
we used restricted receptive fields and a weight sharing mechanism ( i.e. convolutional network ) . in our model ,
images are presented sequentially and the resulting spike waves are propagated through to the @xmath3 layer , where stdp is used to extract diagnostic features . more specifically , the first layer s @xmath0 cells detect bars and edges using gabor filters . here
we used @xmath6 convolutional kernels corresponding to gabor filters with the wavelength of 5 and four different preferred orientations ( @xmath7 ) .
these filters are applied to five scaled versions of the original ) .
hence , for each scaled version of the input image we have four @xmath0 maps ( one for each orientation ) , and overall , there are 4@xmath85 @xmath9 20 maps of @xmath10 cells ( see the @xmath10 maps of fig .
[ model_figure ] ) . evidently , the @xmath0 cells of larger scales detect edges with higher spatial frequencies while the smaller scales extract edges with lower spatial frequencies . indeed , instead of changing the size and spatial frequency of gabor filters
, we are changing the size of input image .
this is a way to implement scale invariance at a low computational cost .
each @xmath10 cell emits a spike with a latency that is inversely proportional to the absolute value of the convolution .
thus , the more strongly a cell is stimulated the earlier it fires ( intensity - to - latency conversion , as observed experimentally @xcite ) . to increase the sparsity at a given scale and location ( corresponding to one cortical column ) , only the spike corresponding to the best matching orientation
is propagated ( i.e. a winner - take - all inhibition is employed ) . in other word , for each position in the four @xmath0 orientation maps of a given scale , the @xmath0 cell with highest convolution value emits a spike and prevents the other three @xmath0 cells from firing . for each @xmath10 map , there is a corresponding @xmath11 map .
each @xmath11 cell propagates the first spike emitted by the @xmath10 cells in a @xmath12 square neighborhood of the @xmath10 map which corresponds to one specific orientation and one scale ( see the @xmath11 maps of fig . [ model_figure ] ) .
@xmath11 cells thus execute a maximum operation over the @xmath10 cells with the same preferred feature across a portion of the visual field , which is a biologically plausible way to gain local shift invariance @xcite .
the overlap between the afferents of two adjacent @xmath2 cells is just one @xmath0 row , hence a subsampling over the @xmath0 maps is done by the @xmath2 layers as well .
therefore , each @xmath2 map has @xmath13 fewer cells than the corresponding @xmath0 map .
@xmath14 features correspond to intermediate - complexity visual features which are optimum for object classification @xcite .
each @xmath14 feature has a prototype @xmath14 cell ( specified by a @xmath11-@xmath14 synaptic weight matrix ) , which is a weighted combination of bars ( @xmath11 cells ) with different orientations in a @xmath15 square neighborhood .
each prototype @xmath14 cell is retinotopically duplicated in the five scale maps ( i.e. weight - sharing is used ) . within those maps
, the @xmath14 cells can integrate spikes only from the four @xmath11 maps of their corresponding processing scales .
this way , a given @xmath14 feature is simultaneously explored in all positions and scales ( see @xmath14 maps of fig .
[ model_figure ] with same feature prototype but in different processing scales specified by different colors ) .
indeed , duplicated cells in all positions of all scale maps integrate the spike train in parallel and compete with each other . the first duplicate reaching its threshold , if any , is the winner .
the winner fires and prevents the other duplicated cells in all other positions and scales from firing through a winner - take - all inhibition mechanism .
then , for each prototype , the winner @xmath14 cell triggers the unsupervised stdp rule and its weight matrix is updated .
the changes in its weights are applied over all other duplicate cells in different positions and scales ( weight sharing mechanism ) .
this allows the system to learn frequent patterns , independently of their position and size in the training images .
the learning process begins with @xmath14 features initialized by random numbers drawn from a normal distribution with mean @xmath16 and std @xmath17 , and the threshold of all @xmath14 cells is set to 64 ( @xmath18 ) . through the learning process , a local inhibition between different @xmath14 prototype cells is used to prevent the convergence of different @xmath14 prototypes to similar features : when a cell fires at a given position and scale , it prevents all the other cells ( independently of their preferred prototype ) from firing later at the same scale and within a neighborhood around the firing position .
thus , the cell population self - organizes , each cell trying to learn a distinct pattern so as to cover the whole variability of the inputs .
moreover , we applied a k - winner - take - all strategy in @xmath14 layer to ensure that at most two cells can fire for each processing scale .
this mechanism , only used in the learning phase , helps the cells to learn patterns with different real sizes . without it ,
there is a natural bias toward small " patterns ( i.e. , large scales ) , simply because corresponding maps are larger , and so likeliness of firing with random weights at the beginning of the stdp process is higher . a simplified version of stdp is used to learn the @xmath19 weights as follows : @xmath20 where @xmath21 and @xmath22 respectively refer to the index of post- and presynaptic neurons , @xmath23 and @xmath24 are the corresponding spike times , @xmath25 is the synaptic weight modification , and @xmath26 and @xmath27 are two parameters specifying the learning rate .
note that the exact time difference between two spikes ( @xmath28 ) does not affect the weight change , but only its sign is considered .
these simplifications are equivalent to assuming that the intensity - to - latency conversion of @xmath10 cells compresses the whole spike wave in a relatively short time interval ( say , @xmath29 ms ) , so that all presynaptic spikes necessarily fall close to the postsynaptic spike time , and the time lags are negligible .
the multiplicative term @xmath30 ensures the weights remain in the range [ 0,1 ] and maintains all synapses in an excitatory mode .
the learning phase starts by @xmath31 which is multiplied by 2 after each 400 postsynaptic spikes up to a maximum value of @xmath32 . a fixed @xmath33 ratio ( -4/3 ) is used .
this allows us to speed up the convergence of @xmath14 features as the learning progresses .
initiation of the learning phase with high learning rates would lead to erratic results . for each @xmath14 prototype
, a @xmath34 cell propagates the first spike emitted by the corresponding @xmath14 cells over all positions and processing scales , leading to the global shift- and scale - invariant cells ( see the @xmath34 layer of fig .
[ model_figure ] ) .
to study the robustness of our model with respect to different transformations such as scale and viewpoint , we evaluated it on the _ 3d - object _ and _ eth-80 _ datasets .
the 3d - object is provided by savarese et al . at cvglab , stanford university @xcite .
this dataset contains 10 different object classes : bicycle , car , cellphone , head , iron , monitor , mouse , shoe , stapler , and toaster .
there are about 10 different instances for each object class .
the object instances are photographed in about 72 different conditions : eight view angles , three distances ( scales ) , and three different tilts .
the images are not segmented and the objects are located in different backgrounds ( the background changes even for different conditions of the same object instance ) .
figure [ objects ] presents some examples of objects in this dataset .
the eth-80 dataset includes 80 3d objects in eight different object categories including apple , car , toy cow , cup , toy dog , toy horse , pear , and tomato .
each object is photographed in 41 viewpoints with different view angles and different tilts .
figure s1 in supplementary information provides some examples of objects in this dataset from different viewpoints . for both datasets ,
five instances of each object category are selected for the training set to be used in the learning phase .
the remaining instances constitute the testing set which is not seen during the learning phase , but is used afterward to evaluate the recognition performance .
this standard cross - validation procedure allows to measure the generalization ability of the model beyond the specific training examples .
note that for 3d - object dataset , the original size of all images were preserved , while the images of eth-80 dataset are resized to @xmath35 pixels in height while preserving the aspect ratio .
the images of both datasets were converted to grayscale values . [ cols="^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] in an other experiment , we analyzed the class dependency of the @xmath4 features for our model . to this end , the 50 most informative features , when classifying a specific class against all the other classes ,
are selected by employing the mutual information technique . in other words , for each class , we selected those 50 features which have the highest activity for samples of that class and have less activity for other classes .
afterwards , the number of common features among the informative features of each pair of classes are computed as provided in table [ table_example2 ] . on average , there are only about 5.4 common features between pairs of classes .
although there are some common features between any two classes , their co - occurrence with the other features help the classifier to separate them from each other . in this way , our model can represent various object classes with a relatively small number of features .
indeed , exploiting the intermediate complexity features , which are not common in all classes and are not very rare , can help the classifier to discriminate instances of different classes @xcite . in a previous study @xcite
, it has been shown that using the hmax model with random dot patterns in the @xmath14 layer can reach a reasonable performance , comparable to the one obtained with random patches cropped from the training images .
it seems that this is due to the dependency of hmax to the application of a powerful classifier .
indeed , the use of both random dot or randomly selected patches transform the images into a complex and nested feature space and it is the classifier which looks for a complex signature to separate object classes .
the deficiencies emerge when the classification problem gets harder ( such as invariant or multiclass object recognition problems ) and then even a powerful classifier is not able to discriminate the classes @xcite . here
, we show that the superiority of our model is due to the informative feature extraction through a bio - inspired learning rule . to this end , we have compared the performances on 3d - object dataset obtained with random features versus stdp features , as well as a very simple classifier versus svm . to generate random features ,
we have set the weight matrix of each @xmath14 feature of our model with random values .
first , we have computed the mean and standard deviation ( std ) ( @xmath36 ) of the number of active ( nonzero ) weights in the features learned by stdp .
second , for each random feature , the number of active weights , @xmath37 , is computed by generating a random number based on the obtained mean and std .
finally , a random feature is constructed by uniformly distributing the @xmath37 randomly generated values in the weight matrix .
in addition , we designed a simple classifier comprised of several one - versus - one classifiers .
for each binary classifier , two subset of @xmath34 features with high occurrence probabilities in one of the two classes are selected . in more details , to select suitable features for the first class , the occurrence probabilities of @xmath34 features in this class are divided by the corresponding occurrence probabilities in the second class .
then , a feature is selected if this ratio is higher than a threshold .
the optimum threshold value is computed by a trial and error search in which the performance over the training samples is maximized . to assign a class label to the input test sample
, we performed an inner product on the feature value and feature probability vectors .
finally , the class with the highest probability is reported to the combined classifier .
the combined classifier selects the winner class based on a simple majority voting . for 500 random features , using the svm and the simple classifier , our model reached classification performances of 71% and 21% on average , respectively .
whereas , for the learned @xmath14 features , both the svm and simple classifiers attained reasonable performances of 96% and 79% , respectively . based on these results , it can be concluded that the features obtained through the bio - inspired unsupervised learning projects the objects into an easily separable space , while the feature extraction by selection of random patches ( drawn from the training images ) or by generation of random patterns leads to a complex object representation .
position and scale invariance in our model are built - in , thanks to weight sharing and scaling process .
conversely , view - invariance must be obtained through the learning process . here , we used all images of five object instances from each category ( varied in all dimensions ) to learn the @xmath14 visual features , while images of all other object instances of each category were used to test the network .
hence , the model was exposed to all possible variations during the learning to gain view - invariance . moreover , near or opposite views of the same object shares some features which are suitable for invariant object recognition .
for instance , consider the overall shape of a head , or close views of a bike wheel which could be a complete circle or an ellipse .
regarding the fact that stdp tends to learn more frequent features in different images , different views of an object could be invariantly represented based on more common features .
our model appears to be the best choice when dealing with few object classes , but huge variations in view points .
as pointed out in previous studies , both hmax and deepconvnet models could not handle these variations perfectly @xcite .
conversely , our model is not appropriate to handle many classes , which requires thousands of features , like in the imagenet contest , because its time complexity is roughly in @xmath38 , where @xmath37 is the number of features ( briefly : since the number of firing neurons per image is limited , if the number of features is doubled , reaching convergence will take roughly twice as many images , and the processing time for each of them will be doubled as well ) .
for example , extracting 4096 features in our model , the same number of features in deepconvnet , would take about 67 times it took us to extract 500 .
however , parallel implementation of our algorithm could speed - up the computation time by several orders of magnitude @xcite . even in this case
, we do not expect to outperform the deepconvnet model on the imagenet database , since only the shape similarities are taken into account in our model and the other cues such as color or texture are ignored .
importantly , our algorithm has a natural tendency to learn salient contrasted regions @xcite , which is desirable as these are typically the most informative @xcite .
most of our @xmath4 features turned out to be class - specific , and we could guess what they represent by doing the reconstructions ( see fig . [ features ] and fig .
since each feature results from averaging multiple input images , the specificity of each instance is averaged out , leading to class archetypes .
consequently , good classification results can be obtained using only a few features , or even using ` simple ' decision rules like feature counts @xcite and majority voting ( here ) , as opposed to a ` smart classifier ' such as svm .
there are some similarities between stdp - based feature learning , and non - negative matrix factorization @xcite , as first intuited in @xcite , and later demonstrated mathematically in @xcite . within both approaches ,
objects are represented as ( positive ) sums of their parts , and the parts are learned by detecting consistently co - active input units .
our model could be efficiently implemented in hardware , for example using address event representation ( aer ) @xcite . with aer ,
the spikes are carried as addresses of sending or receiving neurons on a digital bus .
time ` represents itself ' as the asynchronous occurrence of the event @xcite .
thus the use of stdp will lead to a system which effectively becomes more and more reactive , in addition to becoming more and more selective .
furthermore , since biological hardware is known to be incredibly slow , simulations could run several order of magnitude faster than real time @xcite .
as mentioned earlier , the primate visual system extracts the rough content of an image in about 100ms .
we thus speculate that some dedicated hardware will be able to do the same in the order of a millisecond or less .
recent computational @xcite , psychophysical @xcite , and fmri @xcite experiments demonstrate that the informative intermediate complexity features are optimal for object categorization tasks .
but the possible neural mechanisms to extract such features remain largely unknown .
the hmax model ignores these learning mechanisms and imprints its features with random crops from the training images @xcite , or even uses random filters @xcite .
most individual features are thus not very informative , yet in some cases , a ` smart ' classifier such as svm can efficiently separate the high - dimensional vectors of population responses .
many other models use supervised learning rules @xcite , sometimes reaching impressive performance on natural image classification tasks @xcite .
the main drawback of these supervised methods , however , is that learning is slow and requires numerous labeled samples ( e.g. , about 1 million in @xcite ) , because of the credit assignment problem @xcite .
this contrasts with humans who can generalize efficiently from just a few training examples @xcite .
we avoid the credit assignment problem by keeping the @xmath4 features fixed when training the final classifier ( that being said , fine - tuning them for a given classification problem would probably increase the performance of our model @xcite ; we will test this in future studies ) .
even if the efficiency of such hybrid unsupervised - supervised learning schemes has been known for a long time , few alternative unsupervised learning algorithms have been shown to be able to extract complex and high - level visual features ( see @xcite ) .
finding better representational learning algorithms is thus an important direction for future research and seeking for inspiration in the biological visual systems is likely to be fruitful @xcite .
we suggest here that the physiological mechanism known as stdp is an appealing start point . considering the time relation among the incoming inputs is an important aspect of spiking neural networks .
this property is critical to promote the existing models from static vision to continuous vision @xcite .
a prominent example is the trace learning rule @xcite , suggesting that the invariant object representation in ventral visual system is instructed by the implicit temporal contiguity of vision .
also , in various motion processing and action recognition problems @xcite , the important information lies in the appearance timing of input features .
our model has this potential to be extended for continuous and dynamic vision something that we will further explore .
to date , various bio - inspired network architectures for object recognition have been introduced , but the learning mechanism of biological visual systems has been neglected . in this paper , we demonstrate that the association of both bio - inspired network architecture and learning rule results in a robust object recognition system .
the stdp - based feature learning , used in our model , extracts frequent diagnostic and class specific features that are robust to deformations in stimulus appearance .
it has previously been shown that the trivial models can not tolerate the identity preserving transformations such as changes in view , scale , and position . to study the behavior of our model confronted with these difficulties , we have tested our model over two challenging invariant object recognition databases which includes instances of 10 different object classes photographed in different views , scales , and tilts .
the categorization performances indicate that our model is able to robustly recognize objects in such a severe situation .
in addition , several analytical techniques have been employed to prove that the main contribution to this success is provided by the unsupervised stdp feature learning , not by the classifier . using representational dissimilarity matrix ,
we have shown that the representation of input images in @xmath34 layer are more similar for within - category and dissimilar for between - category objects . in this way , as confirmed by the hierarchical clustering , the objects with the same category are represented in neighboring regions of @xmath34 feature space . hence , even if using a simple classifier , our model is able to reach an acceptable performance , while the random features fail .
we would like to thank mr .
majid changi ashtiani at the math computing center of ipm ( http://math.ipm.ac.ir/mcc ) for letting us to perform some parts of the calculations on their computing cluster .
we also thank dr .
reza ebrahimpour for his helpful discussions and suggestions .
10 url # 1`#1`urlprefixhref # 1#2#2 # 1#1 s. thorpe , d. fize , c. marlot , et al . ,
speed of processing in the human visual system , nature 381 ( 6582 ) ( 1996 ) 520522 .
i. biederman , recognition - by - components : a theory of human image understanding .
, psychological review 94 ( 2 ) ( 1987 ) 115 .
j. j. dicarlo , d. zoccolan , n. c. rust , how does the brain solve visual object recognition ? , neuron 73 ( 3 ) ( 2012 ) 415434 .
p. lennie , j. a. movshon , coding of color and form in the geniculostriate visual pathway ( invited review ) , journal of the optical society of america a 22 ( 10 ) ( 2005 ) 20132033 .
a. s. nandy , t. o. sharpee , j. h. reynolds , j. f. mitchell , the fine structure of shape tuning in area v4 , neuron 78 ( 6 ) ( 2013 ) 11021115 .
k. tanaka , h .- a .
saito , y. fukada , m. moriya , coding visual images of objects in the inferotemporal cortex of the macaque monkey , journal of neurophysiology 66 ( 1 ) ( 1991 ) 170189 .
c. p. hung , g. kreiman , t. poggio , j. j. dicarlo , fast readout of object identity from macaque inferior temporal cortex , science 310 ( 5749 ) ( 2005 ) 863866 .
n. c. rust , j. j. dicarlo , selectivity and tolerance ( `` invariance '' ) both increase as visual information propagates from cortical area v4 to it , journal of neuroscience 30 ( 39 ) ( 2010 ) 1297812995 .
h. liu , y. agam , j. r. madsen , g. kreiman , timing , timing , timing : fast decoding of object information from intracranial field potentials in human visual cortex , neuron 62 ( 2 ) ( 2009 ) 281290 .
w. a. freiwald , d. y. tsao , functional compartmentalization and viewpoint generalization within the macaque face - processing system , science 330 ( 6005 ) ( 2010 ) 845851 .
f. anselmi , j. z. leibo , l. rosasco , j. mutch , a. tacchetti , t. poggio , unsupervised learning of invariant representations with low sample complexity : the magic of sensory cortex or a new framework for machine learning ? , arxiv preprint arxiv:1311.4158 ( 2014 ) 123 .
k. fukushima , neocognitron : a self organizing neural network model for a mechanism of pattern recognition unaffected by shift in position . ,
biological cybernetics 36 ( 4 ) ( 1980 ) 193202 .
y. lecun , y. bengio , convolutional networks for images , speech , and time series , in : m. a. arbib ( ed . ) , the handbook of brain theory and neural networks , cambridge , ma : mit press , 1998 , pp .
255258 .
t. serre , l. wolf , s. bileschi , m. riesenhuber , t. poggio , robust object recognition with cortex - like mechanisms , ieee transactions on pattern analysis machine intelligence 29 ( 3 ) ( 2007 ) 411426 . http://dx.doi.org/10.1109/tpami.2007.56 [ ] .
h. lee , r. grosse , r. ranganath , a. y. ng , convolutional deep belief networks for scalable unsupervised learning of hierarchical representations , proceedings of the 26th annual international conference on machine learning ( icml ) ( 2009 ) 18http://dx.doi.org/10.1145/1553374.1553453 [ ] .
a. krizhevsky , i. sutskever , g. hinton , imagenet classification with deep convolutional neural networks . , in : neural information processing systems ( nips ) , lake tahoe , nevada , 2012 , pp .
q. v. le , building high - level features using large scale unsupervised learning , in : 2013 ieee international conference on acoustics , speech and signal processing , ieee , 2013 , pp .
http://dx.doi.org/10.1109/icassp.2013.6639343 [ ] .
g. m. ghose , learning in mammalian sensory cortex .
, current opinion in neurobiology 14 ( 4 ) ( 2004 ) 513518 .
z. kourtzi , j. j. dicarlo , learning and neural plasticity in visual object recognition .
, current opinion in neurobiology 16 ( 2 ) ( 2006 ) 152158 . http://dx.doi.org/10.1016/j.conb.2006.03.012 [ ] . c. d. meliza , y. dan , receptive - field modification in rat visual cortex induced by paired visual stimulation and single - cell spiking .
, neuron 49 ( 2 ) ( 2006 ) 183189 . http://dx.doi.org/10.1016/j.neuron.2005.12.009 [ ] .
s. huang , c. rozas , m. trevino , j. contreras , s. yang , l. song , t. yoshioka , h .- k .
lee , a. kirkwood , associative hebbian synaptic plasticity in primate visual cortex , journal of neuroscience 34 ( 22 ) ( 2014 ) 75757579 .
http://dx.doi.org/10.1523/jneurosci.0983-14.2014 [ ] .
d. e. feldman , the spike - timing dependence of plasticity .
, neuron 75 ( 4 ) ( 2012 ) 556571 . http://dx.doi.org/10.1016/j.neuron.2012.08.001 [ ] .
d. b. t. mcmahon , d. a. leopold , stimulus timing - dependent plasticity in high - level vision .
, current biology 22 ( 4 ) ( 2012 ) 332337 .
http://dx.doi.org/10.1016/j.cub.2012.01.003 [ ] .
t. masquelier , s. j. thorpe , unsupervised learning of visual features through spike timing dependent plasticity .
, plos computational biology 3 ( 2 ) ( 2007 ) e31 .
s. savarese , l. fei - fei , 3d generic object categorization , localization and pose estimation , in : ieee 11th international conference on computer vision ( iccv ) , 2007 , pp .
http://dx.doi.org/10.1109/iccv.2007.4408987 [ ] .
b. pepik , m. stark , p. gehler , b. schiele , multi - view priors for learning detectors from sparse viewpoint data , in : international conference on learning representations ( iclr ) , banff , ab , canada , 2014 , pp .
b. leibe , b. schiele , analyzing appearance and contour based methods for object categorization , in : computer vision and pattern recognition , 2003 . proceedings .
2003 ieee computer society conference on , vol . 2 , ieee , 2003 , pp .
ii409 .
n. pinto , d. d. cox , j. j. dicarlo , why is real - world visual object recognition hard ? , plos computational biology 4 ( 1 ) ( 2008 )
http://dx.doi.org/10.1371/journal.pcbi.0040027 [ ] .
n. pinto , y. barhomi , d. d. cox , j. j. dicarlo , comparing state - of - the - art visual features on invariant object recognition tasks , in : ieee workshop on applications of computer vision ( wacv ) , ieee , 2011 , pp .
463470 .
m. ghodrati , a. farzmahdi , k. rajaei , r. ebrahimpour , s .- m .
khaligh - razavi , feedforward object - vision models only tolerate small image variations compared to human , frontiers in computational neuroscience 8 ( 2014 ) 74 .
j. pohjalainen , o. rsnen , s. kadioglu , feature selection methods and their combinations in high - dimensional classification of speaker likability , intelligibility and personality traits , computer speech & language .
n. kriegeskorte , m. mur , p. bandettini , representational similarity analysis connecting the branches of systems neuroscience , frontiers in systems neuroscience 2 ( 4 ) .
f. murtagh , p. contreras , algorithms for hierarchical clustering : an overview , wiley interdisciplinary reviews : data mining and knowledge discovery 2 ( 1 ) ( 2012 ) 8697 .
s. j. thorpe , m. imbert , biological constraints on connectionist modelling , in : connectionism in perspective , amsterdam : elsevier , 1989 , pp .
s. celebrini , s. thorpe , y. trotter , m. imbert , dynamics of orientation coding in area v1 of the awake primate . ,
visual neuroscience 10 ( 5 ) ( 1993 ) 811825 .
d. g. albrecht , w. s. geisler , r. a. frazor , a. m. crane , visual cortex neurons of monkeys and cats : temporal dynamics of the contrast response function , journal of neurophysiology 88 ( 2 ) ( 2002 ) 888913 .
o. shriki , a. kohn , m. shamir , fast coding of orientation in primary visual cortex . ,
plos computational biology 8 ( 6 ) ( 2012 ) e1002536 .
m. riesenhuber , t. poggio , hierarchical models of object recognition in cortex .
, nature neuroscience 2 ( 11 ) ( 1999 ) 10191025 .
g. a. rousselet , s. j. thorpe , m. fabre - thorpe , taking the max from neuronal responses .
, trends cogn sci 7 ( 3 ) ( 2003 ) 99102 .
s. ullman , m. vidal - naquet , e. sali , visual features of intermediate complexity and their use in classification .
, nature neuroscience 5 ( 7 ) ( 2002 ) 682687 .
j. mutch , u. knoblich , t. poggio , cns : a gpu - based framework for simulating cortically - organized networks , tech .
mit - csail - tr-2010 - 013 / cbcl-286 , massachusetts institute of technology , cambridge , ma ( february 2010 ) .
y. jia , e. shelhamer , j. donahue , s. karayev , j. long , r. girshick , s. guadarrama , t. darrell , caffe : convolutional architecture for fast feature embedding , arxiv preprint arxiv:1408.5093 .
d. d. cox , t. dean , neural networks and neuroscience - inspired computer vision , current biology 24 ( 18 ) ( 2014 ) r921r929 .
j. z. leibo , j. mutch , s. ullman , t. poggio , from primal templates to invariant recognition , mit - csail - tr-2010 - 057 , cbcl-293 , massachusetts institute of technology , cambridge , ma .
b. lemoine , a. s. maida , gpu facilitated unsupervised visual feature acquisition in spiking neural networks , in : neural networks ( ijcnn ) , the 2013 international joint conference on , ieee , 2013 , pp .
r. vanrullen , s. j. thorpe , rate coding versus temporal order coding : what the retinal ganglion cells tell the visual cortex . , neural comput 13 ( 6 ) ( 2001 ) 12551283 .
d. d. lee , h. s. seung , learning the parts of objects by non - negative matrix factorization . , nature 401 ( 6755 ) ( 1999 ) 788791 .
t. masquelier , s. j. thorpe , learning to recognize objects using waves of spikes and spike timing - dependent plasticity , in : the 2010 international joint conference on neural networks ( ijcnn ) , 2010 , pp . 18 .
http://dx.doi.org/10.1109/ijcnn.2010.5596934 [ ] .
k. carlson , m. richert , n. dutt , j. krichmar , biologically plausible models of homeostasis and stdp : stability and learning in spiking neural networks , in : the 2013 international joint conference on neural networks ( ijcnn ) , 2013 , pp .
18 . c. zamarreo ramos , l. a. camuas mesa , j. a. prez - carrasco , t. masquelier , t. serrano - gotarredona , b. linares - barranco , on spike - timing - dependent - plasticity , memristive devices , and building a self - learning visual cortex , frontiers in neuroscience 5 ( march ) ( 2011 ) 22 .
o. bichler , d. querlioz , s. j. thorpe , j .-
bourgoin , c. gamrat , extraction of temporally correlated features from dynamic vision sensors with spike - timing - dependent plasticity . , neural networks : the official journal of the international neural network society 32 ( 2012 ) 33948 . http://dx.doi.org/10.1016/j.neunet.2012.02.022 [ ] .
t. dorta , m. zapata , j. madrenas , g. snchez , aer - srt : scalable spike distribution by means of synchronous serial ring topology address event representation , neurocomputing 171 ( 2016 ) 16841690 .
c. diaz , g. sanchez , g. duchen , m. nakano , h. perez , an efficient hardware implementation of a novel unary spiking neural network multiplier with variable dendritic delays , neurocomputing .
m. sivilotti , wiring considerations in analog vlsi systems with application to field - programmable networks , ph.d .
thesis , comput .
, california inst .
technol . , pasadena , ca ( 1991 ) . t. serrano - gotarredona , t. masquelier , t. prodromakis , g. indiveri , b. linares - barranco , stdp and stdp variations with memristors for spiking neuromorphic learning systems . ,
frontiers in neuroscience 7 ( february ) ( 2013 ) 2 .
http://dx.doi.org/10.3389/fnins.2013.00002 [ ] .
a. harel , s. ullman , d. harari , s. bentin , basic - level categorization of intermediate complexity fragments reveals top - down effects of expertise in visual perception , journal of vision 11 ( 8) ( 2011 ) 18 .
y. lerner , b. epshtein , s. ullman , r. malach , class information predicts activation by object fragments in human object areas , journal of cognitive neuroscience 20 ( 7 ) ( 2008 ) 11891206 .
t. serre , a. oliva , t. poggio , a feedforward architecture accounts for rapid categorization .
, proc natl acad sci u s a 104 ( 15 ) ( 2007 ) 64246429 .
http://dx.doi.org/10.1073/pnas.0700622104 [ ] .
d. l. k. yamins , h. hong , c. f. cadieu , e. a. solomon , d. seibert , j. j. dicarlo , performance - optimized hierarchical models predict neural responses in higher visual cortex . , proceedings of the national academy of sciences of the united states of americahttp://dx.doi.org/10.1073/pnas.1403112111 [ ] .
e. t. rolls , g. deco , computational neuroscience of vision , oxford university press , oxford , uk , 2002 .
m. a. ranzato , f. j. huang , y .- l .
boureau , y. lecun , unsupervised learning of invariant feature hierarchies with applications to object recognition , in : ieee conference on computer vision and pattern recognition ( cvpr ) , 2007 , pp .
http://dx.doi.org/10.1109/cvpr.2007.383157 [ ] .
h. goh , n. thome , m. cord , j .- h .
lim , learning deep hierarchical visual feature coding , neural networks and learning systems , ieee transactions on pp ( 99 ) ( 2014 ) 11 .
http://dx.doi.org/10.1109/tnnls.2014.2307532 [ ] .
t. masquelier , relative spike time coding and stdp - based orientation selectivity in the early visual system in natural continuous and saccadic vision : a computational model . , journal of computational neuroscience 32 ( 3 ) ( 2012 ) 42541 . http://dx.doi.org/10.1007/s10827-011-0361-9 [ ] .
p. fldik , learning invariance from transformation sequences , neural computation 3 ( 1991 ) 194200 .
escobar , g. s. masson , t. vieville , p. kornprobst , action recognition using a bio - inspired feedforward spiking network , international journal of computer vision 82 ( 3 ) ( 2009 ) 284301 .
here we provide the results of feature analysis techniques such as rdm and hierarchical clustering on eth-80 dataset for both hmax and our model .
some sample images of eth-80 dataset are shown in fig .
[ eth_samples ] . in fig .
[ rdms_supliment_tims ] and fig .
[ rdms_supliment_hmax ] the rdms of @xmath39 features of our model and hmax in eight view angels are presented , respectively .
it can be seen that our model can better represent classes with high shape similarities such as tomato , apple , and pear or cow , horse , and dog with respect to the hmax model .
also , the hierarchical clustering of whole training data based on their representations on feature spaces of our model and hmax are demonstrated in fig .
[ cluster_tim_eth80 ] and fig.[cluster_hmax_eth80 ] , respectively . as for the 3d - object dataset
, hmax feature extraction leads to a nested representation of different object classes which causes a poor classification accuracy . here
again a huge number of images which belong to different classes are assigned to a large cluster with lower than 0.14 internal dissimilarities .
on the other hand , our model has distributed images of different classes in different regions of @xmath39 feature space .
note that the largest cluster of our model includes the instances of tomato , apple , and pear classes which their shapes are so similar . | retinal image of surrounding objects varies tremendously due to the changes in position , size , pose , illumination condition , background context , occlusion , noise , and nonrigid deformations . but
despite these huge variations , our visual system is able to invariantly recognize any object in just a fraction of a second . to date , various computational models have been proposed to mimic the hierarchical processing of the ventral visual pathway , with limited success . here , we show that the association of both biologically inspired network architecture and learning rule significantly improves the models performance when facing challenging invariant object recognition problems .
our model is an asynchronous feedforward spiking neural network .
when the network is presented with natural images , the neurons in the entry layers detect edges , and the most activated ones fire first , while neurons in higher layers are equipped with spike timing - dependent plasticity .
these neurons progressively become selective to intermediate complexity visual features appropriate for object categorization .
the model is evaluated on _ 3d - object _ and _ eth-80 _ datasets which are two benchmarks for invariant object recognition , and is shown to outperform state - of - the - art models , including deepconvnet and hmax .
this demonstrates its ability to accurately recognize different instances of multiple object classes even under various appearance conditions ( different views , scales , tilts , and backgrounds ) .
several statistical analysis techniques are used to show that our model extracts class specific and highly informative features .
* keywords : * view - invariant object recognition , visual cortex , stdp , spiking neurons , temporal coding |
the singular value decomposition of a real @xmath0 matrix @xmath1 is a factorization of the form @xmath2 , where @xmath3 is a @xmath4 orthogonal matrix , @xmath5 is a @xmath0 rectangular diagonal matrix with non - negative real numbers on the diagonal , and @xmath6 is an @xmath7 orthogonal matrix .
the diagonal entries of @xmath5 are known as the _ singular values _ of @xmath1 .
the @xmath8 columns of @xmath3 are the _ left - singular vectors _ of @xmath1 , while the @xmath9 columns of @xmath10 are the _ right - singular vectors _ of @xmath1 . if @xmath1 is symmetric , the singular values are given by the absolute value of the eigenvalues , and the singular vectors are just the eigenvectors of @xmath1 . here , and in the sequel , whenever we write _ singular vectors _ , the reader is free to interpret this as left - singular vectors or right - singular vectors provided the same choice is made throughout the paper .
consider a real ( deterministic ) @xmath0 matrix @xmath1 with singular values @xmath11 and corresponding singular vectors @xmath12 we will call @xmath1 the data matrix .
in general , the vector @xmath13 is not unique .
however , if @xmath14 has multiplicity one , then @xmath13 is determined up to sign .
an important problem in statistics and numerical analysis is to compute the first @xmath15 singular values and vectors of @xmath1 . in particular , the largest few singular values and corresponding singular vectors
are typically the most important . among others
, this problem lies at the heart of principal component analysis ( pca ) , which has a very wide range of applications ( for many examples , see @xcite and the references therein ) and in the closely related low rank approximation procedure often used in theoretical computer science and combinatorics . in application ,
@xmath16 are typically large and @xmath15 is small , often a fixed constant .
a problem of fundamental importance in quantitative science ( including pure and applied mathematics , statistics , engineering , and computer science ) is to estimate how a small perturbation to the data effects the spectrum .
this problem has been discussed in virtually every text book on quantitative linear algebra and numerical analysis ( see , for instance , @xcite ) .
a basic model is as follows . instead of @xmath1
, one needs to work with @xmath17 , where @xmath18 represents the perturbation matrix .
let @xmath19 denote the singular values of @xmath17 with corresponding singular vectors @xmath20 .
we consider the following natural questions
. is @xmath21 a good approximation of @xmath13 ?
[ quest : weyl ] is @xmath22 a good approximation of @xmath14 ?
these two questions are addressed by the davis - kahan - wedin sine theorem and weyl s inequality .
let us begin with the first question in the case when @xmath23 . a canonical way ( coming from the numerical analysis literature ;
see for instance @xcite ) to measure the distance between two unit vectors @xmath24 and @xmath25 is to look at @xmath26 , where @xmath27 is the angle between @xmath24 and @xmath25 taken in @xmath28 $ ] .
it has been observed by numerical analysts ( in the setting where @xmath18 is deterministic ) for quite some time that the key parameter to consider in the bound is the gap ( or separation ) @xmath29 between the first and second singular values of @xmath1 .
the first result in this direction is the famous davis - kahan sine @xmath30 theorem @xcite for hermitian matrices .
the non - hermitian version was proved later by wedin @xcite . throughout the paper
, we use @xmath31 to denote the spectral norm of a matrix @xmath32 .
that is , @xmath31 is the largest singular value of @xmath32 .
[ wedin ] @xmath33 theorem [ wedin ] is trivially true when @xmath34 since sine is always bounded above by one .
in other words , even if the vector @xmath35 is not uniquely determined , the bound is still true for any choice of @xmath35 . on the other hand , when @xmath36 , the proof of theorem [ wedin ] reveals that the vector @xmath35 is uniquely determined up to sign .
theorem [ wedin ] is a simple corollary of ( * ? ? ?
* theorem v.4.4 ) which is originally due to wedin @xcite ; we present a proof below for completeness .
more generally , one can consider approximating the @xmath37-th singular vector @xmath13 or the space spanned by the first @xmath37 singular vectors @xmath38 .
naturally , in these cases , one must consider the gaps @xmath39 question [ quest : weyl ] is addressed by weyl s inequality . in particular ,
weyl s perturbation theorem @xcite gives the following deterministic bound for the singular values ( see ( * ? ? ?
* theorem iv.4.11 ) for a more general perturbation bound due to mirsky @xcite ) .
[ weyl s bound ] [ theorem : weyl ] @xmath40 for more discussions concerning general perturbation bounds , we refer the reader to @xcite and references therein .
we now pause for a moment to prove theorem [ wedin ] . if @xmath34 , the theorem is trivially true since sine is always bounded above by one .
thus , assume @xmath36 . by theorem [ theorem : weyl ] , we have @xmath41 and hence the singular vectors @xmath42 and @xmath35 are uniquely determined up to sign . by another application of theorem [ theorem : weyl ] , we obtain @xmath43 rearranging the inequalities
, we have @xmath44 therefore , by ( * ? ? ?
* theorem v.4.4 ) , we conclude that @xmath45 and the proof is complete .
let us now focus on the matrices @xmath1 and @xmath18 .
it has become common practice to assume that the perturbation matrix @xmath18 is random .
furthermore , researchers have observed that data matrices are usually not arbitrary .
they often possess certain structural properties . among these properties , one of the most frequently seen is having low rank ( see , for instance , @xcite and references therein ) .
the goal in this paper is to show that in this situation , one can significantly improve classical results like theorems [ wedin ] and [ theorem : weyl ] . to give a quick example ,
let us assume that @xmath1 and @xmath18 are @xmath7 matrices and that the entries of @xmath18 are independent and identically distributed ( iid ) random variables with zero mean , unit variance ( which is just matter of normalization ) , and bounded fourth moment .
it is well known that in this case @xmath46 with high probability .
here we use @xmath47 to denote a term which tends to zero as @xmath9 tends to infinity . ]
* chapter 5 ) .
thus , the above two theorems imply [ wedin - cor ] for any @xmath48 , with probability @xmath49 , @xmath50 and @xmath51 among others , this shows that if one wants accuracy @xmath52 in the first singular vector computation , @xmath1 needs to satisfy @xmath53 we present the results of a numerical simulation for @xmath1 being a @xmath7 matrix of rank 2 when @xmath54 , @xmath55 , and where @xmath18 is a random bernoulli matrix ( its entries are iid random variables that take values @xmath56 with probability @xmath57 ) .
the results , shown in figure [ young ] , turn out to be very different from what predicts .
it is easy to see that for the parameters @xmath54 and @xmath58 , corollary [ wedin - cor ] does not give a useful bound ( since @xmath59 ) .
however , figure [ young ] shows that , with high probability , @xmath60 , which means @xmath35 approximates @xmath42 with a relatively small error . where @xmath1 is a @xmath7 deterministic matrix with rank @xmath61 ( @xmath54 for the figure on top and @xmath62 for the one below ) and the noise @xmath18 is a bernoulli random matrix , evaluated from @xmath63 samples ( top figure ) and @xmath64 samples ( bottom figure ) . in both figures ,
the largest singular value of @xmath1 is taken to be @xmath65.,title="fig:",width=453 ] where @xmath1 is a @xmath7 deterministic matrix with rank @xmath61 ( @xmath54 for the figure on top and @xmath62 for the one below ) and the noise @xmath18 is a bernoulli random matrix , evaluated from @xmath63 samples ( top figure ) and @xmath64 samples ( bottom figure ) . in both figures ,
the largest singular value of @xmath1 is taken to be @xmath65.,title="fig:",width=453 ]
trying to explain the inefficiency of the davis - kahan - wedin bound in the above example , the second author was led to the following intuition . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if @xmath1 has rank @xmath66 , all actions of @xmath1 focus on an @xmath66 dimensional subspace ; intuitively then , @xmath18 must act like an @xmath66 dimensional random matrix rather than an @xmath9 dimensional one . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this means that the _ real dimension _ of the problem is @xmath66 , not @xmath9 . while it is clear that one can not automatically ignore the ( rather wild ) action of @xmath18 outside the range of @xmath1 , this intuition ,
if true , would show that what really matters in or is @xmath66 , the rank of @xmath1 , rather than its size @xmath9 . if this is indeed the case , one may hope to obtain a bound of the form @xmath67 for some constant @xmath68 ( with some possible corrections )
this is much better than when @xmath1 has low rank and explains the phenomenon arising from figure [ young ] . in @xcite ,
the second author managed to prove @xmath69 under certain conditions .
while the right - hand side is quite close to the optimal form in , the main problem here is that in the left - hand side one needs to square the sine function .
the bound for @xmath70 with @xmath71 was done by an inductive argument and was rather complicated .
finally , the problem of estimating the singular values was not addressed at all in @xcite . in this paper , by using an entirely different ( and simpler ) argument
, we are going to remove the unwanted squaring effect .
this enables us to obtain a near optimal improvement of the davis - kahan - wedin theorem .
one can easily extend the proof to give a ( again near optimal ) bound on the angle between two subspaces spanned by the first few singular vectors of @xmath1 and their counterparts of @xmath17 .
( this is the space one often actually cares about in pca and low rank approximation procedures . ) finally , as a co - product , we obtain an improved version of weyl s bound , which also supports our _ real dimension _ intuition .
our results hold under very mild assumptions on @xmath1 and @xmath18 . as a matter of fact , in the strongest results
, we will not even need the entries of @xmath18 to be independent .
as an illustration , let us first state a result in the case that @xmath1 is a @xmath7 matrix and @xmath18 is a bernoulli matrix ( the entries are iid bernoulli random variables , taking values @xmath56 with probability @xmath57 ) . [
theorem : main1 ] let @xmath72 be a @xmath7 bernoulli random matrix and fix @xmath73 .
then there exists constants @xmath74 ( depending only on @xmath52 ) such that the following holds .
let @xmath1 be a @xmath7 matrix with rank @xmath66 satisfying @xmath75 and @xmath76 . then , with probability at least @xmath77 , @xmath78 notice that the assumptions on @xmath18 are normalized ( as we assume that the variance of the entries in @xmath18 is one ) .
if the error entries have variance @xmath79 , then we need to scale accordingly by replacing @xmath80 by @xmath81 ; thus , the assumptions become weaker as @xmath82 decreases . for the singular values ,
a good toy result is the following [ thm : probweyl0 ] let @xmath18 be an @xmath7 bernoulli random matrix and fix @xmath83
. then there exists a constant @xmath84 ( depending only on @xmath52 ) such that the following holds .
let @xmath1 be an @xmath7 matrix with rank @xmath66 satisfying @xmath85 .
then with probability at least @xmath77 @xmath86 it may be useful for the reader to compare these new bounds with the bounds obtained directly from the davis - kahan - wedin sine theorem and weyl s inequality ( see corollary [ wedin - cor ] ) .
both theorems above are corollaries of much more general statements , which we describe in the next sections .
in the literature , there are many models of random matrices .
we can capture almost all natural models by focusing on a common property .
[ def : concentration ] we say the @xmath0 random matrix @xmath18 is @xmath87-concentrated if for all unit vectors @xmath88 , and every @xmath89 , @xmath90 the key parameter is @xmath91 . it is easy to verify the following fact , which asserts that the concentration property is closed under addition . [ fact1 ] if @xmath92 is @xmath87-concentrated and @xmath93 is @xmath94-concentrated , then @xmath95 is @xmath96-concentrated for some @xmath97 depending on @xmath98 .
furthermore , the concentration property guarantees a bound on @xmath99 . a standard net argument ( see lemma [ lemma : net ] ) shows [ fact2 ] if @xmath18 is @xmath87-concentrated then there are constants @xmath100 such that @xmath101 . for readers not familiar with random matrix theory , let us point out why the concentration property is expected to hold for any natural model
. if @xmath18 is random and @xmath24 is fixed , then the vector @xmath102 must look random .
it is well known that in a high dimensional space , a random vector , with very high probability , is nearly orthogonal to any fixed vector .
thus , one expects that very likely , the inner product of @xmath103 and @xmath104 is small . definition [ def : concentration ] is a way to express this observation quantitatively .
it turns out that all random matrices with independent entries satisfying a mild condition have the concentration property .
this class covers virtually all examples one sees in practice .
in particular , lemma [ lemma : bernoulli ] shows that if @xmath18 is a @xmath7 bernoulli random matrix , then @xmath18 is @xmath105-concentrated , and @xmath106 with high probability @xcite .
a convenient feature of the definition is that independence between the entries is not a requirement .
for instance , it is easy to show that a random orthogonal matrix satisfies the concentration property .
we continue the discussion of the @xmath107-concentration property ( definition [ def : concentration ] ) in section [ sec : concentration ] .
let us state an extension of theorem [ theorem : main1 ] .
[ thm : main ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 , and suppose @xmath1 has rank @xmath66 .
then , for any @xmath89 , @xmath110 with probability at least @xmath111 [ remark : boundone ] using fact [ fact2 ] , one can replace @xmath112 on the right - hand side by @xmath113 , which yields that @xmath114 with probability at least @xmath115 however , we prefer to state our theorems in the form of theorem [ thm : main ] , as the bound @xmath116 , in many cases , may not be optimal .
another useful corollary of theorem [ thm : main ] is the following .
for any constant @xmath117 there are constants @xmath118 and @xmath119 such that if @xmath75 , then @xmath120 with probability at least @xmath77
. the first term @xmath121 on the right - hand side corresponds to the conjectured optimal bound .
the second term @xmath122 is necessary .
if @xmath123 , then the intensity of the noise is much stronger than the strongest signal in the data matrix , so @xmath18 would corrupt @xmath1 completely . thus in order to retain crucial information about @xmath1 , it seems necessary to assume @xmath124 . we are not absolutely sure about the necessity of the third term @xmath125 , but under the condition @xmath126 , this term is superior to the davis - kahan - wedin bound @xmath127 .
we are able to extend theorem [ thm : main ] in two different ways .
first , we can bound the angle between @xmath128 and @xmath129 for any index @xmath130 .
second , and more importantly , we can bound the angle between the subspaces spanned by @xmath131 and @xmath132 , respectively . as the projection onto the subspaces spanned by the first few singular vectors ( i.e. low rank approximation ) plays an important role in a vast collection of problems ,
this result potentially has a large number of applications .
we are going to present these two results in the next section . to conclude this section ,
let us mention that related results have been obtained in the case where the random matrix @xmath18 contains gaussian entries . in @xcite , r. wang estimates the non - asymptotic distribution of the singular vectors when the entries of @xmath18 are iid standard normal random variables .
recently , allez and bouchaud have studied the eigenvector dynamics of @xmath17 when @xmath1 is a real symmetric matrix and @xmath18 is a symmetric brownian motion ( that is , @xmath18 is a diffusive matrix process constructed from a family of independent real brownian motions ) @xcite .
our results also seems to have a close tie to the study of spiked covariance matrices , where a different kind of perturbation has been considered ; see @xcite for details
. it would be interesting to find a common generalization for these problems .
first , we consider the problem of approximating the @xmath130-th singular vector @xmath128 for any @xmath130 . in light of the davis - kahan - wedin result and theorem [ thm : main ] , it is natural to consider the gap @xmath133 [ thm : general ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 , and let @xmath134 be an integer .
then , for any @xmath89 , @xmath135 with probability at least @xmath136 in the next theorem , we bound the largest principal angle between @xmath137 for some integer @xmath134 , where @xmath66 is the rank of @xmath1 .
let us recall that if @xmath3 and @xmath10 are two subspaces of the same dimension , then the ( principal ) angle between them is defined as @xmath138 where @xmath139 denotes the orthogonal projection onto subspace @xmath140 .
[ thm : subspace ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 , and let @xmath134 be an integer .
then , for any @xmath89 , @xmath141 with probability at least @xmath142 where @xmath10 and @xmath143 are the @xmath130-dimensional subspaces defined in .
it remains an open question to give an efficient bound for subspaces corresponding to an arbitrary set of singular values .
however , we can use theorem [ thm : subspace ] repeatedly to obtain bounds for the case when one considers a few intervals of singular values . for instance , by applying theorem [ thm : subspace ] twice , we obtain assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 , and let @xmath144 be integers .
then , for any @xmath89 , @xmath145 with probability at least @xmath146 where @xmath147 let @xmath148 for any subspace @xmath140 , let @xmath139 denote the orthogonal projection onto @xmath140 .
it follows that @xmath149 , where @xmath150 denotes the identity matrix . by definition of the subspaces @xmath151
, we have @xmath152 thus , by , we obtain @xmath153 theorem [ thm : subspace ] can now be invoked to bound @xmath154 and @xmath155 , and the claim follows .
finally , let us present the general form of theorem [ thm : probweyl0 ] for singular values .
[ thm : probweyl ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 , and let @xmath134 be an integer .
then , for any @xmath89 , @xmath156 with probability at least @xmath157 and @xmath158 with probability at least @xmath159 notice that the upper bound for @xmath160 given in involves @xmath161 . in many situations , the lower bound in
can be used to provide an upper bound for @xmath161 .
we now briefly give an overview of the paper and discuss some of the key ideas behind the proof of our main results .
for simplicity , let us assume that @xmath1 and @xmath18 are @xmath7 real symmetric matrices .
( in fact , we will symmetrize the problem in section [ sec : prelim ] below . )
let @xmath162 be the eigenvalues of @xmath1 with corresponding ( orthonormal ) eigenvectors @xmath163 .
let @xmath164 be the largest eigenvalue of @xmath17 with corresponding ( unit ) eigenvector @xmath35 .
suppose we wish to bound @xmath165 ( from theorem [ thm : main ] ) .
since @xmath166 it suffices to bound @xmath167 for @xmath168 .
let us consider the case when @xmath169 . in this case
, we have @xmath170 since @xmath171 and @xmath172 , we obtain @xmath173 thus , the problem of bounding @xmath167 reduces to obtaining an upper bound for @xmath174 and a lower bound for the gap @xmath175 .
we will obtain bounds for both of these terms by using the concentration property ( definition [ def : concentration ] ) .
more generally , in section [ sec : prelim ] , we will apply the concentration property to obtain lower bounds for the gaps @xmath176 when @xmath177 , which will hold with high probability .
let us illustrate this by now considering the gap @xmath178 .
indeed , we note that @xmath179 applying the concentration property , we see that @xmath180 with probability at least @xmath181 . as @xmath182
, we in fact observe that @xmath183 thus , if @xmath184 is sufficiently large , we have ( say ) @xmath185 with high probability . in section [ sec : proof ]
, we will again apply the concentration property to obtain upper bounds for terms of the form @xmath186 . at the end of section [ sec :
proof ] , we combine these bounds to complete the proof of theorems [ thm : main ] , [ thm : general ] , [ thm : subspace ] and [ thm : probweyl ] . in section [ sec :
concentration ] , we discuss the @xmath107-concentration property ( definition [ def : concentration ] ) . in particular , we generalize some previous results obtained by the second author in @xcite . finally , in section [
section : app ] , we present some applications of our main results . among others ,
our results seem useful for matrix recovery problems .
the general matrix recovery problem is the following .
@xmath1 is a large matrix .
however , the matrix @xmath1 is unknown to us .
we can only observe its noisy perturbation @xmath17 , or in some cases just a small portion of the perturbation .
our goal is to reconstruct @xmath1 or estimate an important parameter as accurately as possible from this observation .
furthermore , several problems from combinatorics and theoretical computer science can also be formulated in this setting .
special instances of the matrix recovery problem have been investigated by many researchers using spectral techniques and combinatorial arguments in ingenious ways @xcite .
we propose the following simple analysis : if @xmath1 has rank @xmath66 and @xmath187 , then the projection of @xmath17 on the subspace @xmath143 spanned by the first @xmath130 singular vectors of @xmath17 is close to the projection of @xmath17 onto the subspace @xmath10 spanned by the first @xmath130 singular vectors of @xmath1 , as our new results show that @xmath10 and @xmath143 are very close .
moreover , we can also show that the projection of @xmath18 onto @xmath10 is typically small .
thus , by projecting @xmath17 onto @xmath143 , we obtain a good approximation of the rank @xmath130 approximation of @xmath1 . in certain cases ,
we can repeat the above operation a few times to obtain sufficient information to recover @xmath1 completely or to estimate the required parameter with high accuracy and certainty .
in this section , we present some of the preliminary tools we will need to prove theorems [ thm : main ] , [ thm : general ] , [ thm : subspace ] , and [ thm : probweyl ] .
to begin , we define the @xmath188 symmetric block matrices @xmath189 and @xmath190 we will work with the matrices @xmath191 and @xmath192 instead of @xmath1 and @xmath18 .
in particular , the non - zero eigenvalues of @xmath191 are @xmath193 and the eigenvectors are formed from the left and right singular vectors of @xmath1 .
similarly , the non - trivial eigenvalues of @xmath194 are @xmath195 ( some of which may be zero ) and the eigenvectors are formed from the left and right singular vectors of @xmath17 . along these lines , we introduce the following notation , which differs from the notation used above .
the non - zero eigenvalues of @xmath191 will be denoted by @xmath193 with orthonormal eigenvectors @xmath196 , @xmath197 such that @xmath198 let @xmath199 be the orthonormal eigenvectors of @xmath200 corresponding to the @xmath130-largest eigenvalues @xmath201 . in order to prove theorems [ thm : main ] , [ thm : general ]
, [ thm : subspace ] , and [ thm : probweyl ] , it suffices to work with the eigenvectors and eigenvalues of the matrices @xmath191 and @xmath200 .
indeed , proposition [ prop : sine ] will bound the angle between the singular vectors of @xmath1 and @xmath17 by the angle between the corresponding eigenvectors of @xmath191 and @xmath194 .
[ prop : sine ] let @xmath202 and @xmath203 be unit vectors .
let @xmath204 be given by @xmath205 then @xmath206 since @xmath207 , we have @xmath208 thus , @xmath209 and the claim follows
. we now introduce some useful lemmas .
the first lemma below , states that if @xmath18 is @xmath107-concentrated , then @xmath192 is @xmath210-concentrated , for some new constants @xmath211 and @xmath212 .
[ lemma : tilde ] assume that @xmath18 is @xmath213-concentrated for a trio of constants @xmath109 .
let @xmath211 and @xmath212 .
then for all unit vectors @xmath214 , and every @xmath89 , @xmath215 let @xmath216 be unit vectors in @xmath217 , where @xmath202 and @xmath218 .
we note that @xmath219 thus , if any of the vectors @xmath220 are zero , follows immediately from .
assume all the vectors @xmath220 are nonzero .
then @xmath221 thus , by , we have @xmath222 and the proof of the lemma is complete
. we will also consider the spectral norm of @xmath192 .
since @xmath192 is a symmetric matrix whose eigenvalues in absolute value are given by the singular values of @xmath18 , it follows that @xmath223 we introduce @xmath52-nets as a convenient way to discretize a compact set .
let @xmath73 .
a set @xmath224 is an @xmath52-net of a set @xmath225 if for any @xmath226 , there exists @xmath227 such that @xmath228 .
the following estimate for the maximum size of an @xmath52-net of a sphere is well - known ( see for instance @xcite ) .
[ lemma : net ] a unit sphere in @xmath229 dimensions admits an @xmath52-net of size at most @xmath230 lemmas [ lemma : r - norm ] , [ lemma : largest ] , and [ lemma : j - largest ] below are consequences of the concentration property . [ lemma : r - norm ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
let @xmath1 be a @xmath0 matrix with rank @xmath66 .
let @xmath3 be the @xmath231 matrix whose columns are the vectors @xmath232 . then , for any @xmath233 , @xmath234 clearly @xmath235 is a symmetric @xmath236 matrix .
let @xmath237 be the unit sphere in @xmath238 .
let @xmath239 be a @xmath240-net of @xmath237 .
it is easy to verify ( see for instance @xcite ) that for any @xmath236 symmetric matrix @xmath241 , @xmath242 for any fixed @xmath243 , we have @xmath244 by lemma [ lemma : tilde ] . since @xmath245
, we obtain @xmath246 [ lemma : largest ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 .
then , for any @xmath233 , @xmath247 with probability at least @xmath248 . in particular ,
if @xmath249 , then @xmath250 with probability at least @xmath251 . if , in addition , @xmath252 , then @xmath253 for @xmath254 with probability at least @xmath255 .
we observe that @xmath256 by lemma [ lemma : tilde ] , we have @xmath257 for every @xmath233 , and follows .
if @xmath249 , then the bound @xmath250 can be obtained by taking @xmath258 in .
assume @xmath252 . taking @xmath259 in yields @xmath260 for @xmath169 with probability at least @xmath255 .
using the courant minimax principle , lemma [ lemma : largest ] can be generalized to the following .
[ lemma : j - largest ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 , and let @xmath134 be an integer .
then , for any @xmath233 , @xmath261 with probability at least @xmath262 . in particular ,
@xmath263 with probability at least @xmath264 .
in addition , if @xmath265 , then @xmath266 for @xmath267 with probability at least @xmath268 .
it suffices to prove .
indeed , the bound @xmath263 follows from by taking @xmath269 , and follows by taking @xmath270 .
let @xmath237 be the unit sphere in @xmath271 . by the courant minimax principle , @xmath272 thus
, it suffices to show @xmath273 for all @xmath233 .
let @xmath239 be a @xmath240-net of @xmath237 . by lemma [ lemma : net ] , @xmath274 .
we now claim that @xmath275 indeed , fix a realization of @xmath192 .
since @xmath237 is compact , there exists @xmath276 such that @xmath277
. moreover , there exists @xmath243 such that @xmath278 .
clearly the claim is true when @xmath279 ; assume @xmath280 .
then , by the triangle inequality , we have @xmath281 and follows . applying and lemma [ lemma : tilde ] , we have @xmath282 and the proof of the lemma is complete .
we will continually make use of the following simple fact : @xmath283
this section is devoted to theorems [ thm : main ] , [ thm : general ] , [ thm : subspace ] , and [ thm : probweyl ] . to begin , define the subspace @xmath284 let @xmath285 be the orthogonal projection onto @xmath286 .
[ lemma : proj_bound ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 , and let @xmath134 be an integer .
then @xmath287 with probability at least @xmath264
. consider the event @xmath288 by lemma [ lemma : j - largest ] ( or lemma [ lemma : largest ] in the case @xmath289 ) , @xmath290 holds with probability at least @xmath264 .
fix @xmath291 . by multiplying on the left by @xmath292 and on the right by @xmath13
, we obtain @xmath293 since @xmath294 .
thus , on the event @xmath290 , we have @xmath295 we conclude that , on the event @xmath290 , @xmath296 and the proof is complete .
[ lemma : uproj ] assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
suppose @xmath1 has rank @xmath66 , and let @xmath134 be an integer .
define @xmath297 to be the @xmath298 matrix with columns @xmath299 .
then , for any @xmath89 , @xmath300 with probability at least @xmath301 define the event @xmath302 by lemmas [ lemma : r - norm ] , [ lemma : j - largest ] , and [ lemma : proj_bound ] , it follows that @xmath303 fix @xmath291 .
we multiply on the left by @xmath304 and on the right by @xmath13 to obtain @xmath305 we note that @xmath306 and @xmath307 where @xmath308 is the diagonal matrix with the values @xmath309 on the diagonal . for the right - hand side of , we write @xmath310 , where @xmath3 is the matrix with columns @xmath232 and @xmath285 is the orthogonal projection onto @xmath286 .
thus , on the event @xmath290 , we have @xmath311 here we used the fact that @xmath312 is a sub - matrix of @xmath235 and hence @xmath313 combining the above computations and bound yields @xmath314 on the event @xmath290 .
we now consider the entries of the diagonal matrix @xmath315 .
on @xmath290 , we have that , for any @xmath316 , @xmath317 by writing the elements of the vector @xmath318 in component form , it follows that @xmath319 and hence @xmath320 on the event @xmath290 .
since this holds for each @xmath291 , the proof is complete . with lemmas [ lemma : proj_bound ] and [ lemma : uproj ] in hand , we now prove theorems [ thm : main ] , [ thm : general ] , [ thm : subspace ] , and [ thm : probweyl ] . by proposition [ prop :
sine ] , in order to prove theorems [ thm : main ] and [ thm : general ] , it suffices to bound @xmath321 because @xmath322 are formed from the left and right singular vectors of @xmath1 and @xmath17 .
we write @xmath323 where @xmath285 is the orthogonal projection onto @xmath286 .
then @xmath324 applying the bounds obtained from lemmas [ lemma : proj_bound ] and [ lemma : uproj ] ( with @xmath289 ) , we obtain @xmath325 with probability at least @xmath326 we now note that @xmath327 the correct absolute constant in front can now be deduced from the bound above and proposition [ prop : sine ] .
the lower bound on the probability given in can be written in terms of the constants @xmath328 by recalling the definitions of @xmath329 and @xmath330 given in lemma [ lemma : tilde ] . we again write @xmath331 where @xmath285 is the orthogonal projection onto @xmath286 .
then we have that @xmath332 for any @xmath333 , we have that @xmath334 moreover , from lemmas [ lemma : proj_bound ] and [ lemma : uproj ] , we have @xmath335 with probability at least @xmath301 and @xmath336 with probability at least @xmath264 .
the proof of theorem [ thm : general ] is complete by combining the bounds above for @xmath333 .
however , @xmath196 and @xmath337 are formed from the left and right singular vectors of @xmath1 and @xmath17 . to avoid the dependence on both the left and right singular vectors
, one can begin with and consider only the coordinates of @xmath128 which correspond to the left ( alternatively right ) singular vectors . by then following the proof for only these coordinates
, one can bound the left ( right ) singular vectors by terms which only depend on the previous left ( right ) singular vectors . ] .
as in the proof of theorem [ thm : main ] , the correct constant factor in front can be deduced from proposition [ prop : sine ] .
define the subspaces @xmath338 by proposition [ prop : sine ] , it suffices to bound @xmath339 .
let @xmath340 be the orthogonal projection onto @xmath341 . by lemmas [ lemma : proj_bound ] and [ lemma : uproj ]
, it follows that @xmath342 with probability at least @xmath343 on the event where holds , we have @xmath344 by the triangle inequality and the cauchy - schwarz inequality .
thus , by , we conclude that @xmath345 on the event where holds .
the claim now follows from proposition [ prop : sine ] .
the lower bound follows from lemma [ lemma : j - largest ] ; it remains to prove .
let @xmath3 be the @xmath231 matrix whose columns are given by the vectors @xmath232 , and recall that @xmath285 is the orthogonal projection onto @xmath286 .
let @xmath237 denote the unit sphere in @xmath346 . then for @xmath291
, we multiply on the left by @xmath347 and on the right by @xmath13 to obtain @xmath348 here we used and the fact that @xmath349 . therefore
, we have the deterministic bound @xmath350 by the cauchy - schwarz inequality , it follows that @xmath351 by the courant minimax principle , we have @xmath352 thus , it suffices to show that @xmath353 with probability at least @xmath354 .
we decompose @xmath355 and obtain @xmath356 thus , by lemma [ lemma : r - norm ] and , we have @xmath357 with probability at least @xmath354 , and the proof is complete .
in this section , we give examples of random matrix models satisfying definition [ def : concentration ] . [ lemma : bernoulli ] there exists a constant @xmath358 such that the following holds .
let @xmath18 be a random @xmath7 bernoulli matrix .
then @xmath359 and for any fixed unit vectors @xmath360 and positive number @xmath361 , @xmath362 the bounds in lemma [ lemma : bernoulli ] also hold for the case where the noise is gaussian ( instead of bernoulli ) . indeed , when the entries of @xmath18 are iid standard normal random variables , @xmath363 has the standard normal distribution .
the first bound is a corollary of a general concentration result from @xcite .
it can also be proved directly using a net argument .
the second bound follows from martingale different sequence inequality @xcite ; see also @xcite for a direct proof with a more generous constant .
we now verify the @xmath107-concentration property for slightly more general random matrix models .
we will discuss these matrix models further in section [ section : app ] . in the lemmas below
, we consider both the case where @xmath18 is a real symmetric random matrix with independent entries and when @xmath18 is a non - symmetric random matrix with independent entries .
[ lemma : conc - sym ] let @xmath364 be a @xmath7 real symmetric random matrix where @xmath365 is a collection of independent random variables each with mean zero .
further assume @xmath366 with probability @xmath367 , for some @xmath368 .
then for any fixed unit vectors @xmath360 and every @xmath233 @xmath369 we write @xmath370 as the right side is a sum of independent , bounded random variables , we apply hoeffding s inequality ( ( * ? ? ?
* theorem 2 ) ) to obtain @xmath371 here we used the fact that @xmath372 because @xmath360 are unit vectors .
since each @xmath373 has mean zero , it follows that @xmath374 , and the proof is complete .
[ lemma : conc - nonsym ] let @xmath375 be a @xmath0 real random matrix where @xmath376 is a collection of independent random variables each with mean zero . further assume @xmath377 with probability @xmath367 , for some @xmath368 .
then for any fixed unit vectors @xmath88 , and every @xmath233 @xmath378 the proof of lemma [ lemma : conc - nonsym ] is nearly identical to the proof of lemma [ lemma : conc - sym ] . indeed , follows from hoeffding s inequality since @xmath379 can be the written as the sum of independent random variables ; we omit the details .
many other models of random matrices satisfy definition [ def : concentration ] .
if the entries of @xmath18 are independent and have a rapidly decaying tail , then @xmath18 will be @xmath107-concentrated for some constants @xmath380 .
one can achieve this by standard truncation arguments .
for many arguments of this type , see for instance @xcite . as an example
, we present a concentration result from @xcite when the entries of @xmath18 are iid sub - exponential random variables .
[ lemma : conc - sub ] let @xmath375 be a @xmath0 real random matrix whose entries @xmath381 are iid copies of a sub - exponential random variable @xmath382 with constant @xmath383 , i.e. @xmath384 for all @xmath89 .
assume @xmath382 has mean 0 and variance 1 .
then there are constants @xmath385 ( depending only on @xmath383 ) such that for any fixed unit vectors @xmath88 and any @xmath233 , one has @xmath386 finally , let us point out that the assumption that the entries are independent is not necessary .
as an example , we mention random orthogonal matrices . for another example
, one can consider the elliptic ensembles ; this can be verified using standard truncation and concentration results , see for instance @xcite and ( * ? ? ?
* chapter 5 ) .
the matrix recovery problem is the following : @xmath1 is a large unknown matrix .
we can only observe its noisy image @xmath17 , or in some cases just a small part of it .
we would like to reconstruct @xmath1 or estimate an important parameter as accurately as possible from this observation . consider a deterministic @xmath0 matrix @xmath387
let @xmath388 be a random matrix of the same size whose entries @xmath389 are independent random variables with mean zero and unit variance . for convenience
, we will assume that @xmath390 , for some fixed @xmath391 , with probability @xmath367 .
suppose that we have only partial access to the noisy data @xmath392 .
each entry of this matrix is observed with probability @xmath393 and unobserved with probability @xmath394 for some small @xmath393 . we will write @xmath395 if the entry is not observed .
given this sparse observable data matrix @xmath241 , the task is to reconstruct @xmath1 . the matrix completion problem is a central one in data analysis , and there is a large collection of literature focusing on the low rank case ; see @xcite and references therein .
a representative example here is the netflix problem , where @xmath1 is the matrix of ratings ( the rows are viewers , the columns are movie titles , and entries are ratings ) . in this section , we are going to use our new results to study this problem .
the main novel feature here is that our analysis allows us to approximate _ any given column ( or row ) _ with high probability .
for instance , in the netflix problem , one can figure out the ratings of any given individual , or any given movie . in earlier algorithms we know of
, the approximation was mostly done for the frobenius norm of the whole matrix .
such a result is equivalent to saying that a _ random _ row or column is well approximated , but can not guarantee anything about a specific row or column . without loss of generality , we assume @xmath1 is a square @xmath7 matrix .
the rectangular case follows by applying the analysis below to the matrix @xmath191 defined in .
we assume that @xmath9 is large and asymptotic notation such as @xmath396 will be used under the assumption that @xmath397 .
let @xmath1 be a @xmath7 deterministic matrix with rank @xmath66 where @xmath398 are the singular values with corresponding singular vectors @xmath399 .
let @xmath400 be iid indicator random variables with @xmath401 .
the entries of the sparse matrix @xmath241 can be written as @xmath402 where @xmath403 it is clear that the @xmath404 are independent random variables with mean 0 and variance @xmath405 .
this way , we can write @xmath406 in the form @xmath407 , where @xmath18 is the random matrix with independent entries @xmath408 .
we assume @xmath409 ; in fact , our result works for @xmath393 being a negative power of @xmath9 .
let @xmath187 and consider the subspace @xmath3 spanned by @xmath410 and @xmath10 spanned by @xmath411 , where @xmath412 ( alternatively @xmath413 is the @xmath37-th singular vector of @xmath1 ( alternatively @xmath241 ) .
fix any @xmath414 and consider the @xmath8-th columns of @xmath1 and @xmath17 .
denote them by @xmath415 and @xmath416 , respectively .
we have @xmath417 notice that @xmath418 is efficiently computable given @xmath241 and @xmath393 .
( in fact , we can estimate @xmath393 very well by the density of @xmath241 , so we do nt even need to know @xmath393 . ) in the remaining part of the analysis , we will estimate the three error terms on the right - hand side .
[ lemma : projection ] let @xmath224 be a random vector in @xmath419 whose coordinates @xmath420 are independent random variables with mean 0 , variance at most @xmath79 , and are bounded in absolute value by @xmath367 .
let @xmath421 be a fixed subspace of dimension @xmath229 and @xmath422 be the projection of @xmath224 onto @xmath421 .
then @xmath423 where @xmath424 are absolute constants . the first term @xmath425 is bounded from above by @xmath426 .
the second term has the form @xmath427 , where @xmath428 is the random vector with independent entries , which is the @xmath8-th column of @xmath18 .
notice that entries of @xmath224 are bounded ( in absolute value ) by @xmath429 with probability @xmath367 .
applying lemma [ lemma : projection ] ( with the proper normalization ) , we obtain @xmath430 since @xmath431 . by setting @xmath432 , implies that , for any @xmath433 , @xmath434 with probability at least @xmath435 . to bound @xmath436 , we appeal to theorem [ thm : subspace ] .
assume for a moment that @xmath18 is @xmath108-concentrated for some constants @xmath437 .
let @xmath438 .
then it follows that , for any @xmath433 , @xmath439 with probability at least @xmath440 where @xmath68 is an absolute constant . since @xmath441 it remains to bound @xmath442 .
we first note that @xmath443 . by talagrand s inequality ( see @xcite or ( * ? ? ?
* theorem 2.1.13 ) ) , we have @xmath444 in addition , @xmath445 thus , we conclude that @xmath446 with probability at least @xmath447 . [
theorem : recovery ] assume that @xmath1 has rank @xmath66 and @xmath448 with probability @xmath367 .
assume that @xmath18 is @xmath108-concentrated for a trio of constants @xmath109 .
let @xmath8 be an arbitrary index between @xmath367 and @xmath9 , and let @xmath415 and @xmath449 be the @xmath8-th columns of @xmath1 and @xmath406 .
let @xmath134 be an integer , and let @xmath10 be the subspace spanned by the first @xmath130 singular vectors of @xmath241 .
let @xmath450 be the singular values of @xmath1 .
assume @xmath438 . then , for any @xmath451 , @xmath452 with probability at least @xmath453 where @xmath454 and @xmath68 is an absolute constant .
as this theorem is a bit technical , let us consider a special , simpler case .
assume that all entries of @xmath1 are of order @xmath455 and @xmath456 .
thus , any column @xmath415 has length @xmath457 .
assume furthermore that @xmath458 and @xmath459 for some @xmath73 .
then our analysis yields | matrix perturbation inequalities , such as weyl s theorem ( concerning the singular values ) and the davis - kahan theorem ( concerning the singular vectors ) , play essential roles in quantitative science ; in particular , these bounds have found application in data analysis as well as related areas of engineering and computer science . in many situations ,
the perturbation is assumed to be random , and the original matrix has certain structural properties ( such as having low rank ) .
we show that , in this scenario , classical perturbation results , such as weyl and davis - kahan , can be improved significantly .
we believe many of our new bounds are close to optimal and also discuss some applications . |
the algebraic boundary of a semi - algebraic set is the smallest algebraic variety containing its boundary in the euclidean topology . for a full - dimensional polytope @xmath0
, it is the hyperplane arrangement associated to its facets which has been studied extensively in discrete geometry and complexity theory in linear programming @xcite .
the algebraic boundary of a convex set which is not a polytope has recently been considered in other special cases , most notably the convex hull of a variety by ranestad and sturmfels , cf .
@xcite and @xcite .
this class includes prominent families such as the moment matrices of probability distributions and the highly symmetric orbitopes .
it does not include examples such as hyperbolicity cones and spectrahedra , which have received attention from applications of semi - definite programming in polynomial optimisation , see @xcite and @xcite , and statistics of gaussian graphical models , see @xcite .
first steps towards using the algebraic boundary of a spectrahedron for a complexity analysis of semi - definite programming have been taken by nie , ranestad , and sturmfels @xcite . for semi - definite liftings of convex semi - algebraic sets via lasserre relaxations or theta body construction
, the singularities of the algebraic boundary on the convex set give obstructions , cf .
@xcite , @xcite
. so algebraic boundaries are central objects in applications of algebraic geometry to convex optimisation and statistics . in this paper , we want to consider the class of all sets for which the algebraic boundary is an algebraic hypersurface : convex semi - algebraic sets with non - empty interior .
our goal in this paper is to extend the study of the algebraic boundary of the convex hull of a variety started by ranestad and sturmfels in @xcite and @xcite to general convex semi - algebraic sets .
the most natural point of view in the general setting is via convex duality and its algebraic counterpart in projective algebraic geometry .
the first main theorem generalizes and implies the correspondence between facets of a polytope with vertices of its dual polytope .
let @xmath1 be a compact convex semi - algebraic set with @xmath2 .
let @xmath3 be an irreducible component of the zariski closure of the set of extreme points of its dual convex body .
then the variety dual to @xmath3 is an irreducible component of the algebraic boundary of @xmath4 . for polytopes ,
this theorem is the whole story . in the general semi - algebraic case ,
not every irreducible component of the algebraic boundary of @xmath4 arises in this way , as we will see below .
we study the exceptional cases and give a complete semi - algebraic description of the exceptional families of extreme points in terms of convex duality ( normal cones ) and a computational way of getting a list of potentially exceptional strata from the algebraic boundary of the dual .
this proves an assertion made by sturmfels and uhler in ( * ? ? ?
* proposition 2.4 ) .
the main techniques come from the duality theories in convex and projective algebraic geometry . for an introduction to convex duality
, we refer to barvinok s textbook @xcite .
the duality theory for projective algebraic varieties is developed in several places , e.g. harris @xcite , tevelev @xcite , or gelfand - kapranov - zelevinsky @xcite .
this article is organized as follows : in section [ sec : prelim ] , we introduce the algebraic boundary of a semi - algebraic set and discuss some special features of convex semi - algebraic sets coming from their algebraic boundary .
the section sets the technical foundation for section [ sec : algbound ] , where we prove the main results of this work .
this section is supposed to be introductory . we will fix notation and observe some basic features of convex semi - algebraic sets , their algebraic boundary , and some special features relying on this algebraic structure .
the main results will be proven in the following section .
let @xmath5 be a semi - algebraic set .
the _ algebraic boundary _ of @xmath6 , denoted as @xmath7 , is the zariski closure in @xmath8 of the euclidean boundary of @xmath6 . in this paper
, we fix a subfield @xmath9 of the complex numbers .
the most important choices to have in mind are the reals , the complex numbers or the rationals .
when we say zariski closure , we mean with respect to the @xmath9-zariski topology , i.e. the topology on @xmath10 ( resp .
@xmath11 ) whose closed sets are the algebraic sets defined by polynomials ( resp .
homogeneous polynomials ) with coefficients in @xmath9
. the set @xmath10 ( resp .
@xmath11 ) equipped with the @xmath9-zariski topology is usually denoted @xmath12 ( resp .
@xmath13 ) .
we drop the field @xmath9 in our notation .
the statements in this paper are true over any subfield @xmath9 of the complex numbers given that the semi - algebraic set in consideration can be defined by polynomial inequalities with coefficients in @xmath14 . if we are interested in symbolic computation , we tend to consider semi - algebraic sets defined by polynomial inequalities with coefficients in @xmath15 and take zariski closures in the @xmath15-zariski topology .
we first want to establish that the algebraic boundary of a convex body is a hypersurface .
a subset of @xmath0 is called _ regular _ if it is contained in the closure ( in the euclidean topology ) of its interior . every convex semi - algebraic set with non - empty interior is regular and the complement of a convex semi - algebraic set is also regular . [
lem : algboundhyp ] let @xmath16 be a regular semi - algebraic set and suppose that its complement @xmath17 is also regular and non - empty .
each irreducible component of the algebraic boundary of @xmath6 has codimension @xmath18 in @xmath8 , i.e. @xmath7 is a hypersurface . by bochnak - coste - roy ( * ? ? ?
* proposition 2.8.13 ) , @xmath19 .
conversely , we prove that the boundary @xmath20 of @xmath6 has local dimension @xmath21 at each point @xmath22 : let @xmath23 be a point and take @xmath24 .
then @xmath25 and @xmath26 are non - empty , because both @xmath6 and @xmath17 are regular . applying ( * ? ? ?
* lemma 4.5.2 ) , yields that @xmath27 therefore , all irreducible components of @xmath28 have dimension @xmath21 .
the assumption of @xmath6 being regular can not be dropped in the above lemma .
write @xmath29 $ ] .
let @xmath6 be the union of the unit ball with the first coordinate axis , i.e. @xmath30 .
the algebraic boundary of @xmath6 is the union of the sphere @xmath31 and the line @xmath32 , which is a variety of codimension @xmath18 with a lower dimensional irreducible component . in the above proof of lemma [ lem :
algboundhyp ] , we argue over the field of real numbers .
the algebraic boundary of @xmath6 , where the zariski closure is taken with respect to the @xmath9-zariski topology for a different field @xmath9 , is also a hypersurface .
it is defined by the reduced product of the galois conjugates of the polynomial defining @xmath7 over @xmath33 , whose coefficients are algebraic numbers over @xmath9 .
let @xmath34 be a compact semi - algebraic convex set with non - empty interior .
its algebraic boundary is a hypersurface .
this property characterises the semi - algebraic compact convex sets .
a compact convex set with non - empty interior is semi - algebraic if and only if its algebraic boundary is a hypersurface .
the converse follows from results in semi - algebraic geometry .
namely if the algebraic boundary @xmath35 is an algebraic hypersurface , its complement @xmath36 is a semi - algebraic set and the closed convex set @xmath4 is the closure of the union of finitely many of its connected components .
this is semi - algebraic by bochnak - coste - roy ( * ? ? ?
* proposition 2.2.2 and theorem 2.4.5 ) . by the construction of homogenisation in convexity
, the algebraic boundary of a pointed and closed convex cone relates to the algebraic boundary of a compact base via the notion of affine cones in algebraic geometry .
[ rem : coneproj ] let @xmath34 be a compact semi - algebraic convex set and let @xmath37 be the convex cone over @xmath4 embedded at height @xmath18 , i.e. @xmath38 . since a point @xmath39 lies in the boundary of @xmath40 if and only if @xmath41 is a boundary point of @xmath4 , the affine cone @xmath42 over the algebraic boundary of @xmath4 is a constructible subset of the algebraic boundary of @xmath40 .
more precisely , we mean that @xmath43 , where @xmath44 is the projective closure of @xmath35 with respect to the embedding @xmath45 , @xmath46 . recall that a closed convex cone @xmath34 is called pointed if @xmath47 , i.e. it does not contain a line .
let @xmath48 be a pointed closed semi - algebraic convex cone .
its algebraic boundary is a hypersurface in @xmath49 and an algebraic cone . in particular
, it is the affine cone over its projectivisation in @xmath50 , i.e. @xmath51 we will now take a look at convex duality for semi - algebraic sets . given a compact convex set @xmath34 , we write @xmath52 for the dual convex set .
we use the notation @xmath53 for the set of all regular ( or smooth ) points of an algebraic variety @xmath44 .
[ prop : expextpoi ] let @xmath34 be a compact semi - algebraic convex set with @xmath2 and set @xmath54 . for every @xmath55 , the face supported by @xmath56 is a point .
the set @xmath6 is an open and dense ( in the euclidean topology ) semi - algebraic subset of the set @xmath57 of all supporting hyperplanes to @xmath4 .
if @xmath58 is a supporting hyperplane to @xmath59 at @xmath56 , then @xmath60 and @xmath59 lies in one halfspace defined by @xmath58 .
therefore , @xmath61 lies locally around @xmath56 in one halfspace defined by @xmath58 and so @xmath58 defines the unique tangent hyperplane to @xmath62 at @xmath56 .
now we show that @xmath41 is an extreme point of @xmath4 , exposed by @xmath56 .
suppose @xmath63 with @xmath64 , then @xmath65 and @xmath66 .
since @xmath67 and @xmath68 are , by the same argument as above , also normal vectors to the tangent hyperplane @xmath69 , we conclude @xmath70 .
the same statement is true for convex cones : we denote the dual convex cone to @xmath71 as @xmath72 .
[ cor : expextpoi ] let @xmath71 be a pointed closed semi - algebraic convex cone with non - empty interior and set @xmath73 . for every @xmath55 , the face supported by @xmath56 is an extreme ray of @xmath4 .
the set @xmath6 is open and dense ( in the euclidean topology ) semi - algebraic subset of @xmath74 .
[ exm : expincmax ] ( a ) in the case that @xmath4 is a polytope , the set @xmath6 of regular points of the algebraic boundary is exactly the set of linear functionals exposing extreme points .
indeed , in this case the algebraic boundary of @xmath4 is a union of affine hyperplanes , namely the affine span of its facets .
a point in @xmath75 is a regular point of the algebraic boundary @xmath35 if and only if it lies in the relative interior of a facet , cf .
barvinok ( * ? ? ?
* theorem vi.1.3 ) .
these points expose the vertices of @xmath59 .
+ ( b ) in general , a linear functional @xmath76 exposing an extreme point of @xmath4 does not need to be a regular point of the algebraic boundary of @xmath59 as the following example shows : let @xmath4 be the convex set in the plane defined by the inequalities @xmath77 , @xmath78 and @xmath79 .
consider the extreme point @xmath80 of @xmath4 .
the dual face is the line segment between the vectors @xmath81 and @xmath82 , the normal vectors to the tangent lines to the curves defined by @xmath83 and @xmath84 , which meet transversally in @xmath41 .
indeed , the linear functionals @xmath81 and @xmath82 both expose extreme points ; but they are each intersection points of a line and a quadric in the algebraic boundary of @xmath59 and so they are singular points of @xmath62 . the extreme points ( resp . rays ) of a convex set play an important role for duality .
they will also be essential in a description of the algebraic boundary using the algebraic duality theory .
so we fix the following notation : \(a ) let @xmath1 be a convex semi - algebraic set .
we denote by @xmath85 the zariski closure of the union of all extreme points of @xmath4 in @xmath8 .
+ ( b ) let @xmath71 be a semi - algebraic convex cone .
we write @xmath86 for the zariski closure of the union of all extreme rays of @xmath4 in @xmath49 .
[ rem : extpointssemialg ] ( a ) note that the union of all extreme points of a convex semi - algebraic set is a semi - algebraic set by quantifier elimination because the definition is expressible as a first order formula in the language of ordered rings , cf .
bochnak - coste - roy
* proposition 2.2.4 ) .
therefore , its zariski closure is an algebraic variety whose dimension is equal to the dimension of @xmath87 as a semi - algebraic set , cf .
bochnak - coste - roy
* proposition 2.8.2 ) . of course
, the same is true for convex cones and the zariski closure of the union of all extreme rays .
+ ( b ) note that @xmath86 is an algebraic cone . in particular , we have @xmath88 let @xmath34 be a compact semi - algebraic convex set with @xmath2 . for a general extreme point @xmath89
there is a supporting hyperplane @xmath90 exposing the face @xmath41 and a semi - algebraic neighbourhood @xmath91 of @xmath92 in @xmath57 such that every @xmath93 supports @xmath4 in an extreme point @xmath94 and all @xmath94 lie on the same irreducible component of @xmath85 as @xmath41 . by general we mean in this context that the statement is true for all points in a dense ( in the zariski topology ) semi - algebraic subset of @xmath85 . by straszewicz s theorem ( see rockafellar ( * ? ?
* theorem 18.6 ) ) and the curve selection lemma from semi - algebraic geometry ( see bochnak - coste - roy ( * ? ? ?
* theorem 2.5.5 ) ) , a general extreme point is exposed .
let @xmath95 be an exposed extreme point contained in a unique irreducible component @xmath3 of @xmath85 and denote by @xmath96 an exposing linear functional .
let @xmath97 be the irreducible components of @xmath85 labelled such that @xmath98 .
since the sets @xmath99 are closed , they are compact .
now @xmath96 is strictly greater than @xmath100 on @xmath101 for @xmath102 and therefore , there is a neighbourhood @xmath91 in @xmath57 of @xmath96 such that every @xmath93 is still strictly greater than @xmath100 on @xmath103 .
the intersection of this neighbourhood with the semi - algebraic set @xmath6 of linear functionals exposing extreme points , which is open and dense in the euclidean topology by proposition [ prop : expextpoi ] , is non - empty and open in @xmath57 . pick @xmath92 from this open set , then the extreme point @xmath41 exposed by @xmath92 has the claimed properties .
\(a ) again , the above lemma has a simple geometric meaning in the case of polytopes : every extreme point of the polytope is exposed exactly by the relative interior points of the facet of the dual polytope dual to it , again by barvinok ( * ? ?
* theorem vi.1.3 ) .
+ ( b ) in example [ exm : expincmax](b ) , the boundary of the convex set @xmath4 consists of extreme points and a single @xmath18-dimensional face .
so the only linear functional not exposing an extreme point of @xmath4 is the dual face to the edge of @xmath4 , which is @xmath104 . by homogenisation
, we can prove the analogous version of the above lemma for closed and pointed convex cones .
[ cor : expextray ] let @xmath48 be a pointed closed semi - algebraic convex cone with non - empty interior .
let @xmath105 be an extreme ray of @xmath4 such that the line @xmath106 $ ] is a general point of @xmath107 .
let @xmath3 be the irreducible component of @xmath107 with @xmath106\in z$ ] .
then there is a supporting hyperplane @xmath108 exposing @xmath109 and a semi - algebraic neighbourhood @xmath91 of @xmath92 in @xmath74 such that every @xmath93 supports @xmath4 in an extreme ray @xmath110 of @xmath4 contained in the regular locus of @xmath3 , i.e. @xmath111\in{z_{{\rm{reg}}}}$ ] .
the above notion of general now translates into the projective notion , i.e. the statement is true for points in a dense semi - algebraic subset of the semi - algebraic set of extreme rays as a subset of @xmath112 .
in this section , we consider a full - dimensional closed semi - algebraic convex cone @xmath48 which is pointed , i.e. it does not contain a line .
the algebraic boundary of @xmath4 is an algebraic cone . in particular , it is the affine cone over its projectivisation , i.e. @xmath113 .
the dual convex cone is the set @xmath114 i.e. the set of all half spaces containing @xmath4 .
we write @xmath86 for the zariski closure of the union of all extreme rays of @xmath4 in @xmath49 . again , this is an algebraic cone .
this is the technically more convenient language for the algebraic duality theory .
we will deduce the statements for convex bodies by homogenisation .
we now consider projective dual varieties : given an algebraic variety @xmath115 , the dual variety @xmath116 is the zariski closure of the set of all hyperplanes @xmath117\in{({{\mathbb}{p}}^{n})^\ast}$ ] such that @xmath118 contains the tangent space to @xmath44 at some regular point @xmath119 . for computational aspects of projective duality
, we refer to ranestad - sturmfels @xcite and rostalski - sturmfels @xcite .
let @xmath121 be an irreducible component of the algebraic boundary of @xmath4 .
let @xmath122 be a general point and @xmath123 be a supporting hyperplane to @xmath4 at @xmath41 .
we argue similarly to the proof of proposition [ prop : expextpoi ] : since @xmath4 lies in one half - space defined by @xmath118 , so does @xmath124 locally around @xmath41 . therefore , @xmath118 is the tangent hyperplane @xmath125 . now
the tangent hyperplane to @xmath124 at @xmath41 is unique , because @xmath124 has codimension @xmath18 .
so the set of all supporting hyperplanes to @xmath4 at @xmath41 is an extreme ray of the dual convex cone . since @xmath126 is zariski dense in @xmath124 , the hyperplanes tangent to @xmath124 at points @xmath127 are dense in the dual variety to @xmath128 .
[ rem : exdual ] let @xmath129 be an irreducible component . then the dual variety to @xmath130 is a hypersurface in @xmath131 , which follows from the biduality theorem in projective algebraic geometry tevelev ( * ? ?
* theorem 1.12 ) , because @xmath132 can not contain a dense subset of projective linear spaces of dimension @xmath133 .
suppose @xmath132 contained a dense subset of projective linear spaces of dimension @xmath133 , then the set @xmath134 , which is dense in @xmath3 , would contain a zariski dense subset of an affine linear space of dimension at least @xmath135 .
this contradicts the fact that the set of extreme rays @xmath136 does not contain any line segments other than those lying on the rays themselves .
[ thm : dualirrcomp ] let @xmath48 be a pointed closed semi - algebraic convex cone with non - empty interior .
the dual variety to the locus of extreme rays of @xmath4 is contained in the algebraic boundary of the dual convex cone @xmath137 , i.e. @xmath138 more precisely , the dual variety to every irreducible component of @xmath107 is an irreducible component of @xmath139 .
let @xmath140 be an irreducible component of the locus of extreme rays of @xmath4 .
by corollary [ cor : expextray ] , a general extreme ray @xmath106\in { { \mathbb}{p}}z\cap ( { { \mathbb}{p}}\operatorname{exr}(c))$ ] is exposed by @xmath108 and there is a semi - algebraic neighbourhood @xmath91 of @xmath92 in @xmath74 such that every @xmath93 exposes an extreme ray @xmath110 of @xmath4 such that @xmath111\in{({{\mathbb}{p}}z)_{{\rm{reg}}}}$ ] .
the hyperplane @xmath141 is tangent to @xmath132 at @xmath111 $ ] because @xmath132 is locally contained in @xmath4 ; so @xmath142 is a semi - algebraic subset of @xmath143 of full dimension and the claim follows .
[ thm : dualirrcompaffine ] let @xmath34 be a compact convex semi - algebraic set with @xmath2 .
let @xmath3 be an irreducible component of the zariski closure of the set of extreme points of its dual convex body .
then the variety dual to @xmath3 is an irreducible component of the algebraic boundary of @xmath4 .
more precisely , the dual variety to the projective closure @xmath144 of @xmath3 with respect to the embedding @xmath145 , @xmath146 is an irreducible component of the projective closure of @xmath35 with respect to @xmath147 , @xmath146 .
we homogenise the convex body and its dual convex body by embedding both at height @xmath18 to get convex cones @xmath148 and @xmath149 .
the projective closure @xmath144 of the irreducible component @xmath150 with respect to the embedding @xmath145 , @xmath146 is an irreducible component of @xmath151 . by the above theorem [ thm : dualirrcomp ] ,
the dual variety to @xmath144 is an irreducible component of @xmath152 , which is the projective closure of an irreducible component of the algebraic boundary of @xmath4 with respect to the embedding @xmath147 , @xmath146 .
[ rem : twoparabolas ] it does not follow from the biduality theorems in both theories that @xmath154 simply because the biduality theorem in the algebraic context does not in general apply to this situation , since the varieties in question tend to be reducible .
in fact , the mentioned equality does not hold in general , as the following example shows : let @xmath155 be the convex set defined by the inequalities @xmath156 and @xmath157 , see figure [ fig : circleplaneduality ] .
the dual convex body is the convex hull of the set @xmath158 and the point @xmath159 ( it can not be defined by simultaneous polynomial inequalities , i.e. it is not a basic closed semi - algebraic set ) .
its algebraic boundary has three components , namely the circle and the two lines @xmath160 and @xmath161 .
the set of extreme points of @xmath4 is @xmath162 .
so @xmath163 and @xmath164 .
( -5.5,0 ) ( -2.5,0 ) ; ( -4,-1.5 ) ( -4,1.5 ) ; ( -3.4,-0.8 ) ( -3.4,0.8 ) ; ( -3.4,0.8 ) ( -3.4,1.5 ) ; ( -3.4,-0.8 ) ( -3.4,-1.5 ) ; ( -4,0 ) circle(1 ) ; ( -3.4,-1 ) rectangle ( -5,1 ) ; ( -4,0 ) circle(1 ) ; ( -4,0 ) circle(1 ) ; the following statement gives a complete semi - algebraic characterisation of the irreducible subvarieties @xmath165 with the property that @xmath166 is an irreducible component of the algebraic boundary of @xmath137 .
[ thm : dualcond ] let @xmath48 be a pointed closed semi - algebraic convex cone .
let @xmath3 be an irreducible algebraic cone contained in @xmath86 and suppose @xmath134 is zariski dense in @xmath3 .
then the dual variety to @xmath132 is an irreducible component of @xmath167 if and only if the dimension of the normal cone to a general point @xmath168 is equal to the codimension of @xmath3 , i.e. @xmath169 conversely , if @xmath128 is an irreducible component of the algebraic boundary of @xmath137 , then the dual variety to @xmath170 is an irreducible subvariety of @xmath107 , the set @xmath171 is zariski dense in @xmath172 and the above condition on the normal cone is satisfied at a general extreme ray for the affine cone over @xmath172 .
consider the semi - algebraic set @xmath174 defined as @xmath175 this is the set of all tuples @xmath176 , where @xmath41 spans an extreme ray of @xmath4 and is a regular point of @xmath3 and @xmath56 is a supporting hyperplane to @xmath4 at @xmath41 , i.e. the fibre of the projection @xmath177 onto the first factor over a point @xmath41 is the normal cone @xmath178 . since a supporting hyperplane to @xmath4 at @xmath41 is tangent to @xmath3 at @xmath41 , this bihomogeneous semi - algebraic incidence correspondence is naturally contained in the conormal variety @xmath179 of the projectivisation of @xmath3 . now
the image @xmath180 is zariski dense in @xmath143 if and only if @xmath143 is an irreducible component of the projectivisation of the algebraic boundary of @xmath137 .
indeed , @xmath181 and so if it is dense in @xmath143 , we immediately get that @xmath182 is an irreducible component , because @xmath143 is a hypersurface ( cf .
remark [ rem : exdual](b ) ) .
conversely , we have seen in the proof of the above proposition that if @xmath182 is an irreducible component , the unique tangent hyperplane to a general point of @xmath183 spans an extreme ray of @xmath4 , i.e. a general point of @xmath183 is contained in @xmath180 . on the other hand , counting dimensions of @xmath184 as the sum of the dimensions of @xmath3 and the dimension of the fibre over a general point in @xmath187
, we see that @xmath188 if and only if the claimed equality of dimensions @xmath189 holds .
the second part of the statement follows from the first by proposition [ prop : dualalgbound ] .
we want to compare this theorem to the result of ranestad and sturmfels in @xcite : they consider the convex hull of a smooth algebraic variety @xmath115 and make the technical assumption that only finitely many hyperplanes are tangent to the variety @xmath44 in infinitely many points , which is needed for a dimension count in the proof .
we get rid of this technical assumption in the above theorem .
the assumption that the extreme rays are zariski dense in the variety @xmath3 in question , compares best to the ranestad - sturmfels assumption .
it is semi - algebraic in nature .
the corresponding affine statement to theorem [ thm : dualcond ] is the following .
we take projective closures with respect to the same embeddings as in the affine version corollary [ thm : dualirrcompaffine ] of theorem [ thm : dualirrcomp ] above .
[ thm : dualcondaffine ] let @xmath34 be a compact convex semi - algebraic set with @xmath2 .
let @xmath3 be an irreducible subvariety of @xmath85 and suppose @xmath190 is dense in @xmath3 .
then the dual variety to @xmath144 is an irreducible component of @xmath191 if and only if @xmath192 for a general extreme point @xmath193 .
conversely , if @xmath128 is an irreducible component of @xmath62 , then the dual variety to @xmath194 is an irreducible subvariety of @xmath195 , the set @xmath196 is dense in @xmath197 and the condition on the normal cone is satisfied at a general extreme point .
let @xmath198 be a basic closed semi - algebraic convex set with non - empty interior defined by @xmath199 $ ] .
then the algebraic boundary @xmath35 is contained in the variety @xmath200 , where @xmath201 are the irreducible factors of the polynomials @xmath202 .
the irreducible hypersurface @xmath203 is an irreducible component of @xmath35 if and only if @xmath204 is a semi - algebraic set of codimension @xmath18 . by the above corollary [ thm : dualcondaffine ]
, we can equivalently check the following conditions on the dual varieties @xmath205 to the projective closure @xmath206 : the dual convex body is the convex hull of a quartic curve . its algebraic boundary is @xmath210 here , the line @xmath211 is a bitangent to the quartic and the dual variety of the node @xmath212 of the cubic and the quartic is the dual curve to the cubic .
we define @xmath4 using the cubic inequality and additionally either one linear inequality or the two tangents to the branches of @xmath44 in @xmath212 @xmath213 and we see both conditions in action .
first , the dual variety to the affine line @xmath214 is @xmath215 , which is not an extreme point of @xmath59 .
the first condition mentioned above shows , that the line @xmath216 corresponding to the second inequality in the first description is not an irreducible component of @xmath35 . in the second description , the dual variety to the affine line @xmath217 is the point @xmath218 , which is an extreme point of @xmath59 .
the normal cone @xmath219 is @xmath18-dimensional , because the supporting hyperplane is uniquely determined - it is the bitangent @xmath211 to the quartic .
so by the second condition above , the line @xmath220 is not an irreducible component of @xmath35 .
[ cor : algboundlocextpoi][to corollary [ thm : dualcondaffine ] ] let @xmath221 be a compact semi - algebraic set with @xmath2 .
let @xmath222 be an irreducible component such that @xmath223 is not an irreducible component .
if @xmath197 is contained in a bigger irreducible subvariety @xmath224 such that @xmath225 is dense in @xmath3 , then let @xmath224 be an irreducible subvariety . if @xmath228 defines a supporting hyperplane to an extreme point @xmath229 that is an interior point of the semi - algebraic set @xmath230 as a subset of @xmath3 and @xmath231 , then the variety @xmath3 lies locally in one of the half spaces defined by @xmath232 and therefore @xmath232 is tangent to @xmath3 at @xmath233 . in particular , the dimension of the normal cone @xmath234 is bounded by the local codimension of @xmath3 at @xmath233 .
now if @xmath197 is strictly contained in @xmath3 , it can not contain @xmath233 by corollary [ thm : dualcondaffine ] because @xmath235 .
consider the convex hull @xmath4 of the half ball @xmath236 and the circle @xmath237 .
the zariski closure of the extreme points of @xmath4 is the sphere @xmath238 .
every point of the circle @xmath44 is a regular point of @xmath238 and @xmath44 is contained in the algebraic boundary of @xmath239 , because the semi - algebraic set @xmath240 does not have local dimension @xmath135 at the extreme points @xmath241 where @xmath242 .
the algebraic boundary of the dual convex set has three irreducible components , namely the sphere @xmath238 and the dual varieties to the two irreducible components @xmath44 and @xmath243 of @xmath244 . consider the spectrahedron @xmath245 where @xmath246 is the symmetric matrix @xmath247{cccc } 1 & x & 0 & x \\ x & 1 & y & 0 \\ 0 & y & 1 & z \\ x & 0 & z & 1 \end{array}\right),\ ] ] studied by rostalski and sturmfels in ( * ? ? ?
* section 1.1 ) and called pillow .
the zariski closure of the set of extreme points of @xmath248 is defined by the equation @xmath249 , where @xmath250 the algebraic boundary of the dual convex body @xmath251 is the hypersurface @xmath252 computed in rostalski - sturmfels ( * ? ? ?
* section 1.1 , equations 1.7 and 1.8 ) .
the first quartic is the dual variety to the quartic @xmath253 .
the two quadric hypersurfaces are products of linear forms over @xmath33 and they are the dual varieties to the four corners of the pillow , namely @xmath254 , @xmath255 , @xmath256 and @xmath257 .
these four points are extreme points of @xmath248 and singular points of @xmath253 .
+ [ cor : extpointscentral ] let @xmath1 be a compact semi - algebraic convex set with @xmath2 .
every extreme point @xmath41 of @xmath4 is a central point of the dual variety of at least one irreducible component of @xmath191 via @xmath147 , @xmath146 .
a point @xmath41 on a real algebraic variety @xmath115 is called central if @xmath258 has full local dimension around @xmath41 .
equivalently , @xmath259 is central if it is the limit of a sequence regular real points of @xmath44 , cf .
bochnak - coste - roy
* section 7.6 and proposition 10.2.4 ) . by straszewicz s theorem ( * ? ? ?
* theorem 18.6 ) , it suffices to prove , that the statement holds for exposed extreme points because every extreme point is the limit of an exposed one .
so let @xmath41 be an exposed extreme point of @xmath4 and let @xmath260 be the dual face .
because @xmath41 is exposed , the normal cone @xmath261 is @xmath18-dimensional . fix a relative interior point @xmath262 .
let @xmath128 be an irreducible component of @xmath62 on which @xmath56 is a central point and let @xmath263 be a sequence of regular real points converging to @xmath56 in the euclidean topology .
there is a unique ( up to scaling ) linear functional minimising in @xmath264 over @xmath59 , namely @xmath265 with @xmath266 and @xmath267 for all @xmath268 .
since @xmath269 is a sequence in a compact set , there exists a converging subsequence ; without loss of generality , we assume that @xmath270 converges and we call the limit @xmath67 . note that @xmath67 represents a central point of @xmath197 . we know
@xmath271 and @xmath272 so @xmath67 exposes the face @xmath273 of @xmath59 and is therefore equal to @xmath41 by @xmath274 .
a homogeneous polynomial @xmath275 $ ] of degree @xmath276 is called hyperbolic with respect to @xmath277 if @xmath278 and the univariate polynomial @xmath279 $ ] has only real roots for every @xmath280 .
we consider the set @xmath281 which is called the hyperbolicity cone of @xmath282 ( with respect to @xmath283 ) .
it turns out to be a convex cone , cf .
assume that all non - zero points in the boundary of @xmath284 are regular points of @xmath285 .
then by corollary [ cor : algboundlocextpoi ] the algebraic boundary of the dual convex cone is the dual variety to @xmath286 where @xmath287 is the unique irreducible factor of @xmath282 which vanishes on @xmath288 .
the assumption on the hyperbolicity cone being smooth is essential : consider the hyperbolicity cone of @xmath289 $ ] with respect to @xmath290 .
the cubic @xmath291 is singular along the line @xmath292 and the algebraic boundary of the dual convex cone has an additional irreducible component , namely the hyperplane dual to this line because the normal cone has dimension @xmath135 at this extreme ray , see figure [ fig : schleife ] .
let now @xmath284 be any hyperbolicity cone and decompose @xmath293 into its irreducible components @xmath294 .
the dual convex cone @xmath295 is the conic hull of the regular real points of the dual varieties of the irreducible components @xmath205 up to closure , i.e. @xmath296 indeed , the right hand side contains every central point of every variety @xmath297 and by corollary [ cor : extpointscentral ] , this gives one inclusion .
conversely , let @xmath56 be a general real point of @xmath297 for any @xmath298 .
then @xmath56 is tangent to @xmath205 in a regular real point of @xmath299 and by hyperbolicity of @xmath282 , the linear functional has constant sign on the hyperbolicity cone @xmath284 because every line through the hyperbolicity cone intersects every regular real point of @xmath299 with multiplicity @xmath18 , cf .
plaumann - vinzant ( * ? ? ?
* lemma 2.4 ) .
how can we compute these exceptional varieties of extreme points ? given the algebraic boundary of the dual convex set , the following theorem gives an answer . in its statement
, we use an iterated singular locus : the @xmath9-th iterated singular locus of a variety @xmath44 , denoted by @xmath300 , is the singular locus of the @xmath301 iterated singular locus .
the @xmath18-st iterated singular locus is the usual singular locus of @xmath44 .
[ thm : comp ] let @xmath34 be a compact semi - algebraic convex set with @xmath302 and suppose that every point @xmath76 is a regular point on every irreducible component of @xmath303 containing it .
let @xmath150 be an irreducible subvariety such that @xmath304 is an irreducible component of @xmath305 .
if @xmath306 , then @xmath3 is an irreducible component of @xmath62 .
if @xmath307 , then @xmath3 is an irreducible component of an iterated singular locus , namely it is an irreducible component of one of the varieties @xmath308 .
assume @xmath309 and let @xmath310 be a general point .
since whitney s condition a is satisfied for @xmath311 at @xmath56 for every irreducible component @xmath312 with @xmath313 by bochnak - coste - roy ( * ? ? ?
* theorem 9.7.5 ) , every extreme ray @xmath314 of @xmath315 is tangent to @xmath3 at @xmath56 by corollary [ cor : extpointscentral ] . since the extreme rays of the normal cone @xmath315 span the smallest linear space containing it , the dimension of @xmath3 is bounded from above by @xmath316 .
the assumption that @xmath304 is an irreducible component of @xmath305 implies @xmath317 by corollary [ thm : dualcondaffine ] .
it follows that the tangent space @xmath318 is the lineality space of the convex cone @xmath319 . to show that @xmath3 is an irreducible component of @xmath320 , suppose @xmath321 is an irreducible component with @xmath322 and @xmath323 and
let @xmath310 be a general point with @xmath324 .
then @xmath325 and there is an extreme ray @xmath314 of @xmath326 with @xmath327 and @xmath328 . by corollary [ cor : extpointscentral ] , there is an irreducible component @xmath312 such that @xmath41 is a central point of @xmath329 .
so by assumption , @xmath330 and @xmath331 .
since @xmath332 , the varieties @xmath128 and @xmath44 intersect transverally at @xmath56 .
so @xmath333 and @xmath334 is an irreducible component for some @xmath335 because the multiplicity of a point in @xmath336 in @xmath62 is higher than the multiplicity of a general point on @xmath128 .
induction on the codimension of @xmath3 proofs the theorem .
\(a ) this theorem gives a computational way to get a list of candidates for the dual varieties to irreducible components of the algebraic boundary of @xmath4 , given the algebraic boundary of @xmath59 .
certain of these candidates may fail to contribute an irreducible component due to semi - algebraic constraints . for illustration
, we will apply it to two examples .
+ ( b ) the assumption that all irreducible components of @xmath62 are smooth along the boundary of @xmath59 is used to show that the stratification into iterated singular loci is sufficient in this case . in general
, it may be necessary to refine this stratification such that whitney s condition a is satisfied for all adjacent strata , see example [ exm : whitney ] .
we consider the convex set @xmath155 in the plane defined by the two inequalities @xmath337 and @xmath157 , see figure [ fig : circleplaneduality ] .
its algebraic boundary is the plane curve @xmath338 .
the dual convex body is the convex hull of the set @xmath339 and the point @xmath159 .
its algebraic boundary is the curve @xmath340 .
its three irreducible components are smooth and its singular locus consists of three points , namely @xmath159 and @xmath341 . by the above theorem , a complete list of candidates for the algebraic boundary of @xmath4 are the dual varieties to the circle @xmath342 and the irreducible components of the first iterated singular locus , i.e. the lines dual to the points @xmath159 and @xmath341 .
in fact , the last two points do not contribute an irreducible component to @xmath35 , because the normal cone to @xmath59 at these points is @xmath18-dimensional , cf .
corollary [ thm : dualcondaffine ] .
+ we can also look at it dually and compute the algebraic boundary @xmath62 from the singularities of the algebraic boundary of @xmath4 : the curve @xmath35 is reducible , all components are smooth , and its singular locus consists of two points , namely @xmath343 . both of these points dualize to irreducible components of @xmath62 . as an example in @xmath344-space , consider the convex set @xmath4 defined as the intersection of two affinely smooth cylinders given by the inequalities @xmath337 and @xmath345 .
the algebraic boundary of @xmath4 is the ( reducible ) surface @xmath346 , whose singular locus is a smooth curve of degree @xmath347 , namely the intersection of the two cylinders . since the dual varieties to the cylinders are curves and the iterated singular loci of @xmath35 are this smooth curve of degree @xmath347 or empty , the algebraic boundary of the dual convex body is , by theorem [ thm : comp ] , the dual variety of this curve , which is a surface of degree @xmath348 defined by the polynomial @xmath349 viewed dually , this example is more complicated .
the algebraic boundary of @xmath59 is the surface of degree @xmath348 defined by the above polynomial , which has singularities along the boundary of @xmath59 .
so the above theorem is not applicable in this case but the conclusion is still true and we compute the iterated singular loci for demonstration .
the singular locus of the surface has @xmath347 irreducible components : the dual varieties to the cylinders , which are circles , namely @xmath350 and @xmath351 , a complex conjugate pair of quadrics @xmath352 , and a curve of degree @xmath353 , which we denote by @xmath354 . the second iterated singular locus , which is the singular locus of the union of these @xmath347 irreducible curves , consists of @xmath355 points .
@xmath356 of them are the singular points of @xmath354 and the other @xmath348 points are intersection points of @xmath354 with the complex conjugate pair of quadrics @xmath352 .
the two circles dual to the cylinders intersect the curve @xmath354 only in singular points of the latter .
there are no other intersection points of the irreducible components of @xmath357 . of these @xmath355 points in @xmath358 only @xmath347 are real .
they are @xmath359 and @xmath360 .
now the difficult job is to exclude those varieties that do not contribute irreducible components to the algebraic boundary of @xmath4 .
the dual variety to @xmath62 is only a curve , so it can not be an irreducible component of @xmath35 .
next , we discuss the irreducible components of @xmath357 : the dual varieties to the complex conjugate pair of quadrics can not be an irreducible component of @xmath35 either , because the real points will not be dense in this hypersurface . why the dual variety to the curve @xmath354 is not an irreducible component of @xmath35 is not obvious . of the irreducible components of @xmath358 , the @xmath347 real points must be considered as potential candidates for dual varieties to irreducible components of @xmath35 .
to close , we want to consider an example of a convex set whose algebraic boundary is not smooth along its euclidean boundary and for which the conclusion of the theorem [ thm : comp ] is false .
as remarked above , the stratification into iterated singular loci must be refined to a stratification that is whitney a - regular .
[ exm : whitney ] consider the surface in @xmath361 defined by @xmath362 which is the union of an irreducible cubic and two hyperplanes meeting along the line @xmath363 .
the cubic surface is a rotation of the nodal curve shown in figure [ fig : schleife ] on the left along the @xmath41-axis , so the convex set @xmath4 bounded by the cubic looks like a teardrop .
we consider the extreme point @xmath364 of @xmath4 : the normal cone is two - dimensional and so the dual hyperplane @xmath365 is an irreducible component of the algebraic boundary of @xmath59 .
indeed , the point @xmath282 is a singular point of the cubic that lies on the line @xmath363 , which is an irreducible component of the singular locus of the reducible surface @xmath366 , so @xmath282 can not be found by computing the iterated singular loci of @xmath366 .
we make this discussion relevant by perturbing the above polynomial @xmath367 in such a way that it becomes irreducible and shows the same behaviour : consider the polynomial @xmath368 which is irreducible over @xmath15 .
the surface @xmath369 is the algebraic boundary of a convex set @xmath370 , a perturbation of the teardrop @xmath4 .
convexity of @xmath370 can be checked by writing @xmath68 as a function of @xmath41 and @xmath67 and checking its convexity resp .
concavity using its hessian matrix ( note that @xmath68 only occurs to the power of @xmath135 in @xmath371 ) .
the point @xmath282 is also an extreme point of @xmath370 and the normal cone at @xmath282 relative to @xmath370 is still @xmath135-dimensional . yet
the algebraic boundary of @xmath370 is only singular along the line @xmath363 , which is a smooth curve .
so we do nt find @xmath372 as an irreducible component of an iterated singular locus of @xmath373 .
note that whitney s condition a for @xmath374 is not satisfied at @xmath282 because a hyperplane that is in limiting position for supporting hyperplanes to the teardrop @xmath370 do not contain the line @xmath363 .
refining the stratification of iterated singular loci into a whitney a - regular stratification would detect this special extreme point .
* acknowledgements .
* this work is part of my phd thesis .
i would like to thank my advisor claus scheiderer for his encouragement and support , the studienstiftung des deutschen volkes for their financial and ideal support of my phd project , and the national institute of mathematical sciences in korea , which hosted me when i finished this paper .
jacek bochnak , michel coste , and marie - franoise roy . , volume 36 of _ ergebnisse der mathematik und ihrer grenzgebiete ( 3 ) [ results in mathematics and related areas ( 3)]_. springer - verlag , berlin , 1998 . translated from the 1987 french original , revised by the authors .
kristian ranestad and bernd sturmfels .
the convex hull of a variety . in petter branden , mikael passare , and mihai putinar , editors , _ notions of positivity and the geometry of polynomials _ , trends in mathematics , pages 331344 .
springer , basel , 2011 .
victor vinnikov .
lmi representations of convex semialgebraic sets and determinantal representations of algebraic hypersurfaces : past , present , and future . in _ mathematical methods in systems ,
optimization , and control _ , volume 222 of _ oper .
theory adv .
_ , pages 325349 .
birkhuser / springer basel ag , basel , 2012 . | we study the algebraic boundary of a convex semi - algebraic set via duality in convex and algebraic geometry .
we generalize the correspondence of facets of a polytope with the vertices of the dual polytope to general semi - algebraic convex sets . in this case , exceptional families of extreme points might exist and we characterize them semi - algebraically .
we also give an algorithm to compute a complete list of exceptional families , given the algebraic boundary of the dual convex set . |
protostellar outflows are generally byproducts of star formation in the full range from low- to high - mass star - forming regions @xcite . here
we focus our attention on outflows from intermediate - mass protostars of a few solar masses , with typical mass - loss rates of @xmath0 to a few @xmath1yr@xmath2 @xcite and outflow momentum rates from @xmath3 to several @xmath4kms@xmath2yr@xmath2 @xcite .
intermediate - mass outflows are typically elongated with collimation factors between 1 and 10 @xcite , but recent observations in w75n have revealed an apparently very young , spherical outflow @xcite .
there are two essentially independent mechanisms that can drive protostellar outflows with the help of magnetic fields .
first , the disk material can be accelerated centrifugally and launch a disk wind ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . or second
, the gas in the disk can be lifted by the pressure of the toroidal magnetic field in a magnetic tower flow ( e.g. * ? ? ?
* ; * ? ? ?
based upon the lorentz force , these mechanisms are not mutually exclusive and can well act in concert .
@xcite derived an analytic description able to distinguish both types of driving mechanisms along the outflow region which were also tested with numerical simulations .
additionally , outflows around high - mass stars can be driven by radiation pressure @xcite and ionization feedback @xcite .
however , these types of outflows have a very different morphology and kinematics that are on the lower end of observed outflow properties @xcite .
three - dimensional magnetohydrodynamic models of protostellar outflows @xcite show that magnetically - driven outflows can have strongly varying morphologies , ranging from collimated , elongated outflows to almost spherical bubbles .
@xcite found that collimated outflows are only formed when a nearly keplerian disk is present . in idealized numerical simulations of protostellar collapse ( i.e. simulations without initial turbulence ) keplerian disks
build up only in cases of weak magnetic fields @xcite , whereas keplerian disks naturally arise in simulations that include some initial velocity or density perturbations independent of the magnetic field strength @xcite .
the morphologies of the outflows launched from those turbulence - generated disks are not yet studied in detail .
it seems likely that that the associated outflows will be launched by magnetic pressure gradients as well as by magneto - centrifugal forces and will take on different shapes depending on the environment and evolutionary state of the underlying disk . in more massive initial cloud cores
, @xcite found that magnetic bubbles can also be produced from common disks around multiple systems by fragmentation - induced outflow disruption .
here , spherical , low - velocity outflows are generated by a large - scale pseudodisk or toroid that forms around the central high - mass star , which then becomes gravitationally unstable and fragments .
this fragmentation process destroys any coherent rotation and renders both magnetic launching mechanisms impossible .
hence , as more gas falls onto the disk , the radius at which gravitational instability sets in increases and the outflow stalls inside of this radius .
since the toroid itself grows in radius as well and the outer parts of the disk still rotate coherently , the outflow becomes more spherical in shape .
@xcite have recently reported the discovery of such a large - scale spherical tower flow .
fragmentation - induced outflow disruption thus naturally relates to the idea of fragmentation - induced starvation @xcite . even if individual outflows are launched from the disks around single low- and high - mass protostars ( which are not resolved in the simulations by * ? ? ?
* ) within the young multiple system , it might be difficult to observe collimated outflows due to the mutal interaction of those outflows and the influence of the h ii regions around the massive protostars . in all cases ,
the outflows found in numerical simulations show a complex internal structure where knots are generated and different instabilities occur .
often prominent are corkscrew and helical structures due to kink instability or outflow precession ( e.g. * ? ? ?
* ; * ? ? ?
although those internal structures are commonly observed in jets from low - mass stars ( see e.g. , review by * ? ? ?
* ) , the outflow structure from intermediate- and high - mass stars is more obscure . yet ,
the recent observation of an outflow from the young a - type star hd 163296 by @xcite reveals a double sided corkscrew structure that is interpreted as the internal structure of a disk wind launched from around this star . in this present study
we confront those observations with numerical simulations . to avoid the complexities of feedback and fragmentation in high - mass star formation
, we here focus on outflows from intermediate - mass stars .
we make predictions for how the different outflow morphologies found by @xcite would appear to the observer , in particular focusing on the recent observations by @xcite . in section [ sec : rtmodels ] , we describe the simulation snapshots that were observed following the procedure outlined in section [ sec : synobs ] .
we present the results of our analysis in section [ sec : res ] and compare our findings with observations in section [ sec : compobs ] .
we conclude in section [ sec : concl ] .
the simulations are performed with the astrophysical code flash @xcite using the mhd solver devised by @xcite . to follow the long - term evolution of the protostellar disks and their associated outflows we make use of sink particles @xcite . for details of the numerical methods applied we refer to @xcite . since these simulations to not include radiative feedback , the gas can only heat up through hydrodynamic processes , such as compression and the development of shocks . in the following we analyse the results of two simulations which differ only in the strength of the initial magnetic field .
both simulations start with a 100 @xmath5 molecular cloud core , 0.25 pc in diameter and rotating rigidly around the @xmath6-axis with a rotation frequency of @xmath7 s@xmath2 .
the magnetic field is initially aligned with the rotation axis , i.e. parallel to the @xmath6-axis . in run 26 - 4 the magnetic field strength
is chosen such that the normalized mass - to - flux ratio is @xmath8 and the ratio of rotational to gravitational energy is @xmath9 ( the run number encodes these two fundamental quantities of the initial conditions ) .
hence , the core is magnetically supercritical and the magnetic field does not have a strong impact on the collapse of the core . in the second run 5.2 - 4
, the magnetic field strength is increased by a factor of 5 resulting in a mass - to - flux ratio @xmath10 of 5.2 .
again , we refer to @xcite for more details on the initial conditions . during the collapse of the core in run 26 - 4
, a rotationally supported disk builds up around the first protostar , starting to fragment after @xmath11 2600 yr . in contrast , in run 5.2 - 4 a sub - keplerian disk with strong radial infall motions forms , showing no signs of fragmentation until the end of the simulation .
this is a consequence of the efficient removal of angular momentum from the inner parts of the core by magnetic braking @xcite . on the other hand ,
for run 26 - 4 magnetic braking is too weak so that a keplerian disk can build up . in both simulations a magneto - centrifugally driven protostellar outflow
is launched after the formation of the first sink particle .
the outflow in run 26 - 4 has a well - collimated morphology with a collimation factor of @xmath11 4 by the end of the simulation .
the outflowing gas reaches velocities of up to @xmath1120 km s@xmath2 , well above the escape speed .
in contrast , in run 5.2 - 4 a poorly collimated , almost spherical outflow with relatively low outflow velocities up to @xmath117 km s@xmath2 is formed .
both outflows keep expanding in a roughly self - similar fashion keeping the overall morphological properties . while only a single sink particle forms in run 5.2 - 4 , a small cluster develops in run 26 - 4 .
the mass spectrum in this cluster is , however , by far dominated by the central sink particle in our snapshots .
we analyse three snapshots in total , two snapshots for run 26 - 4 ( elon - a and elon - b ) and one snapshot of run 5.2 - 4 ( spher ) .
the masses of the central sink particles in these snapshots are 2.02@xmath12 , 2.89@xmath12 and 2.29 @xmath12 , respectively .
table [ table : direct ] shows the mass , momentum and kinetic energy of the outflows . the values are determined directly from the simulation data by measuring all the outflowing gas more than 50 au above and below the midplane .
unlike observational measurements , we do not restrict contributions to the summed values to a certain velocity range along a particular line of sight .
the outflow elon - a of run 26 - 4 has an average outflow lobe height of 3200 au and an age of 5000 yr , but during the first 1500 yr the outflow grows only very slowly .
the volume - weighted mean temperature is 60 k ( 41 k mass - weighted ) , averaged spatially over the outflow region .
these larger values compared to the initial temperature ( 20 k ) are primarily due to compressive motion and shock heating .
the maximum outflow speed is 19 km s@xmath2 .
note that this maximal velocity is attained close to the disk and that most of the gas in the outflow is a few km s@xmath2 slower . in snapshot elon - b of run 26 - 4
, the maximal velocity is only 14 km s@xmath2 because disk fragmentation at 55006000 yr reduces the outflow activity .
the outflow is now 10000 yr old ( or 8500 yr , disregarding the initial slow starting phase ) and has an average outflow lobe height of 9600 au , consistent with a linear extrapolation from the previous snapshot with a typical outflow velocity of 14 km s@xmath2 . by comparing the masses and kinematics of elon -
a and the later time elon - b in table [ table : direct ] , we find that the momentum and energy of the outflow do not grow by the same factor as the outflow mass .
in fact , the outflow energy even decreases slightly for the blue - shifted component
. the volume - weighted mean temperature is a bit smaller than before and now has the value 52 k ( 39 k mass - weighted ) .
all of these effects are results of the disk fragmentation .
snapshot spher of run 5.2 - 4 , with an average height of one outflow lobe of 1100 au , is 4000 yr old .
the maximum outflow speed is 6.9 km s@xmath2 , less than half the velocity observed in the other simulation .
the volume - weighted mean temperature is 33 k ( 28 k mass - weighted ) .
it is expected that the gas temperature here is lower compared to the other simulations because the smaller outflow velocities compress the gas to a lesser extent .
run 5.2 - 4 has not been followed for longer times , so that we can not analyse another snapshot from a later stage .
[ cols="<,>,^,^,^,^,^ " , ] + outflow mass @xmath13 , momentum @xmath14 , kinetic energy @xmath15 , luminosity @xmath16 and mass - loss rate @xmath17 as determined from the synthetic co observations .
we use the three - dimensional adaptive - mesh radiative transfer code radmc-3ddullemond / software / radmc-3d/ ] to make synthetic co line observations .
we model the molecular line emission of the @xmath18 transition of the isotopologues @xmath19co , @xmath20co and c@xmath21o with abundances relative to h@xmath22 of @xmath3 , @xmath23 and @xmath24 , respectively @xcite .
the critical density of the co @xmath18 transition is @xmath25@xmath26 .
figure 2 and 7 of @xcite show that the minimum density in the outflow is more than two orders of magnitude higher than this value . because of the high density of the outflow material we can assume local thermodynamic equilibrium ( lte ) .
the einstein coefficients of the transitions were taken from the leiden atomic and molecular database @xcite .
the outputs from radmc-3d were converted into skymaps assuming a distance to source of 128 pc for setting the angular scale and converting the fluxes to jy beam@xmath2 .
noiseless alma observations were simulated using the casa tasks `` simobserve '' and `` simanalyze '' .
simulations were carried out in casa version 4.0.0 @xcite .
we simulated full alma observations of the @xmath18 transition of @xmath27co , @xmath28co and c@xmath21o with a spatial resolution of [email protected]@xmath30 , or 83 au .
we have chosen this beam size because it is an average value for the full alma configurations .
the outflows are observed at an inclination of 30@xmath31 with respect to the disk normal direction , which is a representative outflow orientation @xcite .
the resulting images had spectral resolutions of 390 khz ( @xmath11 0.5 km s@xmath2 ) , with the data cubes for the three isotopologues centered at 230.538 , 220.399 and 219.560 ghz respectively .
the simulated fields of view fit within a single alma band 6 pointing , and we used a total integration time of 4 hours .
cleaning was done non - interactively , using natural weighting , and a threshold of 18 mjy .
this limit was chosen to best represent the emission in the maps , as we did not simulate atmospheric noise .
we calculated outflow properties from the @xmath27co map , after having corrected for the opacity of the line using the less optically thick @xmath28co and c@xmath21o isotopologues .
we used equation 1 of @xcite to derive the optical depth of @xmath27co using the least abundant isotoplogue ( c@xmath21o ) towards the line center , and @xmath28co at higher velocities when the c@xmath21o emission fell below our threshold of 18 mjy . to determine the gas mass in each velocity bin , we first calculated the column density from the flux in the given velocity bin scaled by the average line opacity at that velocity , @xmath32 where @xmath33 is the column density of h@xmath22 and @xmath34 is the partition function for converting the column density in the @xmath35=2 level to the overall level populations assuming lte and using a temperature of 50 k ( a compromise between the temperatures derived above in section [ sec : rtmodels ] ) , @xmath36 is the integrated intensity of the line ( in units of k km s@xmath2 ) , and @xmath37 is the abundance of @xmath27co with respect to h@xmath22 .
the gas mass in each channel was then derived by multiplying the column density by the size of the emitting region to obtain the total number of h@xmath22 molecules .
this was then multiplied by the mass of hydrogen , and the mean molecular weight of 2.3 .
these individual masses were summed for all velocities greater than @xmath382 km s@xmath2 , -2.5,+2 km s@xmath2 , and @xmath382.0 km s@xmath2 for the three simulations ( elon - a , elon - b and spher , respectively ) to find the total mass in each outflow lobe .
these velocities were chosen based on visual inspection of the processed spectra .
the limits were chosen individually for each simulation , and were set where the spectra first appear gaussian . to derive the kinematics of the outflows , we multiplied the gas mass in each channel by the velocity of that channel to derive the outflow momentum ( @xmath39 ) and mechanical energies in the outflows ( @xmath40 ) .
the outflow luminosities and mass - loss rates were obtained by dividing the mechanical energies and masses , respectively , by the kinematically derived ages of the outflows . using the spatial extents of the outflows , and the mass - weighted velocities at the end of the outflows , we determined outflow ages of 2400 , 5000 and 800 yr for elon - a , elon - b and spher .
we note that , especially for spher , this is an underestimate of the true outflow age as we are looking down the outflow cavity instead of across it in the plane of the sky .
the derived outflow masses and kinematics are presented in table [ table : co ] .
we note that atmospheric noise is not included in our images .
the uncertainty in our simulated observation comes from the uv coverage and integration time only .
we find typical errors of the order @xmath41 for @xmath13 , @xmath41 km s@xmath2 for @xmath14 , @xmath42erg for @xmath15 , @xmath43 for @xmath16 and @xmath44yr@xmath2 for @xmath17 .
these quantities reflect the noise levels in the synthetic observations which have been propogated through in quadrature .
we used a beam uncertainty of 0.25@xmath30 and velocity uncertainty of 0.25 km s@xmath2 , which represent one third of our spatial and one half of our spectral resolution , respectively .
if we were to take atmospheric noise into account , the alma sensitivity calculator expects we would have a noise level of 6 mjy beam@xmath2 , instead of the measured root - mean - square noise level of 1.3 mjy beam@xmath2 .
when cleaning the data , we used a threshold based on 6 mjy beam@xmath2 .
the first moment and channel maps for elon - a , elon - b and spher are shown in figures [ fig : elona ] , [ fig : elonb ] and [ fig : spher ] , respectively .
figure [ fig : elona ] illustrates that the outflow velocities closer to the star are generally larger .
this is because the material at a location near the star was launched at a later time than material further away . at these later times , the star is more massive and the keplerian velocity greater , increasing the outflow velocity . in general , the velocities in the first moment map well represent the average velocities in the outflow . in figure [
fig : elonb ] , the outflow of elon - b is shown at a larger scale than elon - a .
the first moment map displays a very prominent helical structure .
this helix is the result of an mhd instability occuring for axisymmetric jets , the so - called kink instability ( e.g. * ? ? ?
* ; * ? ? ?
the kink instability describes a helical ( @xmath45 ) displacement of the jet from the symmetry axis without any distortion of the jet profile .
the instability is stabilised by the magnetic field of the jet .
the development of such a kink ( or helical ) instability over time was studied numerically by @xcite who find a successive growth over time .
here we find that the instability starts to grow significantly only after the time at which snapshot elon - a is taken , which is why it is seen more prominently in elon - b . in the maps of elon - a ,
there are velocity structures in the individual channel maps which may be hinting at the existence of a tightly wound rotating structure within the outflow ( see figure [ fig : elona_zoom ] ) , however it is only at later times ( in elon - b ) , that the helical structure in the outflow becomes identifiable in the first moment map .
the bow shock of the outflow in elon - b is visible as a low - velocity shell around the tips of the two outflow lobes in figure [ fig : elonb ] .
the velocities and flux densities measured from spher ( figure [ fig : spher ] ) are much smaller than those obtained from elon - a and elon - b .
this is not surprising since the velocity component along the line of sight is much smaller for the spherical outflow than for the elongated one .
quantitatively , the outflow velocities in spher are a factor of a few smaller than the maser spots in w75n @xcite , on the other hand the size of spher is also larger by a similar factor .
@xcite speculated that the spherical outflow seen in snapshot spher might be a transient feature because the outflow velocity is so small that the outflow could fall back onto the disk . with time , a small keplerian disk around the central star could then build up and make the launching of a collimated outflow possible .
if this scenario is true , then spherical outflows around intermediate - size stars of a size much larger than spher should not be observed .
however , an outflow similar to the one in w75n has been observed around a massive star in cepheus a @xcite . a source in hl tauri that is likely more evolved than the @xcite simulations
is surrounded by a spherical bubble as well @xcite . these observations might suggest that spherical outflows can occur repeatedly during the disk evolution , and not only in an initial transient phase .
since figure [ fig : spher ] does not resemble a typical protostellar outflow , we think that it might be useful as a reference for observers in case they find a similarly looking object . ) .
the number inside each panel of the channel maps is the line - of - sight velocity in kms@xmath2 . ]
the outflow mass and kinematics derived from the simulated observations ( see table [ table : co ] ) are directly comparable to those derived from the modelled outflows themselves ( see table [ table : direct ] ) . the values for mass , momentum and energy are mostly within a factor of two of each other .
the majority of the directly measured values are smaller than the observed ones .
the origin of this behaviour is unclear , and we have seen the opposite trend in previous work @xcite .
the luminosity and mass - loss rate are slightly less accurate because the observationally determined outflow ages are generally less than half of the true values measured directly from the simulations . for elon - a and elon - b the mass measurements
are most consistent , while the measurements of the outflow momentum and energy show larger deviations . for snapshot spher ,
the mass and kinematics derived from the simulated observations are most uncertain because of the spherical outflow morphology .
although we have made every attempt to recover all of the flux in the observed maps and thoroughly cover the uv plane , there may be some missing or enhanced structure which is biasing our derived masses and kinematics .
uncertainties also arise in the derivation of the optical depth of the lines which could bias our derivations , which then over- or underestimates the correction factor in equation .
we used standard observational methods for deriving the outflow masses and kinematics from the simulated observations . that our results are so strikingly similar to the masses and kinematics derived from the models themselves is a testament to the robustness of the methods used .
the helical structure in snapshot elon - b can be compared to recent alma observations of a disk wind around the herbig ae star hd 163296 @xcite .
we show the first moment co map of hd 163296 in figure [ fig : obs ] . for reference ,
the hco@xmath46 emission from the disk is also shown as blue and red contours .
hd 163296 is located at a distance of 122 pc , and therefore the angular scales in figure [ fig : obs ] are directly comparable to those in figure [ fig : elonb ] .
one can see that the spatial scale of the observed helically twisted arcs fits very nicely to our model prediction .
in fact , the kink of the helix is at the same distance from the central star .
the velocities of the hd 163296 wind are , however , much faster than those seen in elon - b ( averaging at 18.6 km s@xmath2 from the source rest velocity ) .
the limits of the observations mean that the large - scale morphology of this wind has yet to be observed , but it is possible that the observed structures in hd 163296 do wind up to a spiral on larger scales .
even though hd 163296 is about 4 myr old , we compare it to elon - b ( where the protostar is about 10 kyr old ) because observations of hd 163296 show for the first time the corkscrew structure that our simulations predict .
map of hd 163296 .
the blue and red contours display the 15 , 20 and 25 times root - mean - square noise ( 20 mjy beam@xmath2 ) of the blue ( -2 to 5.5 km s@xmath2 ) and red ( 6 to 12 km s@xmath2 ) hco@xmath46 @xmath47 emission from the disk .
the brown contours show 4 , 6 , and 8 times the root - mean - square noise in the co @xmath48 emission ( 2.6 mjy beam@xmath2 ) .
the velocities listed in the colour bar are lsr velocities .
the rest velocity of the source is 7 km s@xmath2 .
the blue and red dashed lines delineate the helical structure mentioned in the text . ]
if we degrade the resolution and sensitivity in the maps of elon - b , we would still be able to see velocity gradients within the outflow , such as in elon - a .
@xcite have reported velocity gradients in co observations of hh 797 with the sma .
they report velocity differences of 2 km s@xmath2 over distances of 1000 au , which are roughly consistent with the velocity gradient in elon - a .
alma , with more than 10 times the linear resolution of the sma , will likely see more than a velocity gradient in each lobe of hh 797 , but the precessing gas itself .
the rotation of the protostellar outflow can be seen at larger inclination angles .
figure [ fig : edgeona ] and [ fig : edgeonb ] show co observations of elon - a and elon - b , respectively , at an inclination of 80@xmath31 . we have chosen a small deviation from 90@xmath31 since observations exactly edge - on are very unlikely .
these maps show velocity gradients across the outflow that are clear evidence for outflow rotation .
additionaly , there are velocity gradients along the outflow axis .
these secondary gradients are caused by the growth of the keplerian velocity with time as well as the complex gas motion along the helical structure , exactly the same reasons as for the 30@xmath31 maps .
similar gradients have been found in co observations of an outflow in cb 26 with the pdbi by @xcite . in their study of 16 class 0 sources
, @xcite found that in addition to the outflow motions from these sources , in more than half of them ( 11 ) , there were extra velocity gradients more than 45@xmath31 from the outflow direction .
they suggested that this was either due to contamination by infall , or due to rotation of the outflowing gas itself .
however , our models elon - a and elon - b also show significant velocity gradients not aligned with the outflow axis .
these gradients are created by the complex dynamics of the outflowing gas and by shocked gas in the outflow , not by infall . o peak intensity . ]
o peak intensity . ] in hd 163296 , the outflow can be traced to the disk from which it is launched .
there are other recent alma observations of protostellar outflows in which this connection has not been made .
@xcite have imaged an outflow around a young massive star in orion - kl in sio .
since sio is a shock tracer and the observations do not have many resolution elements across the outflow , we can not expect to see a helical structure .
@xcite have reported alma observations of one of the most energetic and luminous outflows in the milky way , g331.512 - 0.103 .
this outflow is also not well resolved , and our outflows are certainly much weaker .
@xcite have presented alma co observations of the hh 46/47 molecular outflow .
this outflow is much smaller in mass , momentum and energy than any of our outflows , and the associated star appears to be a little bit smaller , too .
however , there appears to be no evidence for a helical structure in the outflow .
we have presented synthetic alma observations of intermediate - mass outflows , derived from the self - consistent magnetohydrodynamic protostellar collapse calculations of @xcite .
we find generally good agreement between outflow properties measured from co lines and the simulation data .
elongated outflows are much easier to detect than spherical outflows because of their larger line - of - sight velocities .
however , the dearth of evidence for spherical outflows could also be explained on statistical grounds if they are just transient objects .
more observations of spherical outflows and simulations that follow the disk and outflow evolution for a longer period are necessary to settle the question .
edge - on views of the elongated outflows show velocity gradients consistent with observations .
we find a helical structure in the co maps that is caused by an instability during the outflow launching .
this helix is already present in the early outflow phases ( @xmath115000 yr ) but becomes very prominent at later times ( @xmath1110000 yr ) .
we speculate that the recent observation of a disk wind in hd 163296 is the first instance of the detection of such a helix , which needs to be backed up by follow - up studies .
we thank the anonymous referee for useful comments that helped to improve the paper . t.p .
acknowledges financial support through snf grant 200020137896 and a forschungskredit of the university of zrich , grant no .
fk-13 - 112 . d.s . acknowledges funding from the _ deutsche forschungsgemeinschaft _ dfg via the grant ba 3706/1 - 3 within the priority program spp 1573 _ physics of the interstellar medium_. r.b .
acknowledges funding from the dfg via the grant ba 3706/1 - 1 .
r.s.k . acknowledges funding from the dfg via grants kl 1358/14 - 1 as part of the spp 1573 as well as via the collaborative research project sbb 811 _ the milky way system _ in subprojects b1 , b2 , and b4 .
the flash code was in part developed by the doe - supported alliances center for astrophysical thermonuclear flashes ( asci ) at the university of chicago . | the formation of stars is usually accompanied by the launching of protostellar outflows .
observations with the atacama large millimetre / sub - millimetre array ( alma ) will soon revolutionalise our understanding of the morphologies and kinematics of these objects . in this paper , we present synthetic alma observations of protostellar outflows based on numerical magnetohydrodynamic collapse simulations .
we find significant velocity gradients in our outflow models and a very prominent helical structure within the outflows .
we speculate that the disk wind found in the alma science verification data of hd 163296 presents a first instance of such an observation . |
thin layers of fluid on a solid substrate display surprisingly rich dynamics , due to the interplay of forces at many lengthscales @xcite .
much progress has been achieved on the study of thin fluid film systems .
they are mostly well characterized by the lubrication approximation of the navier - stokes equations .
this elegant approximate formalism allows for tractable analysis of a wide range of fluid dynamics problems on many lengthscales , such as liquids spreading on flat surfaces @xcite , inclined surfaces @xcite , spin coating applications @xcite , dam breaks @xcite and geophysical @xcite contexts : `` thin '' here means that the height @xmath0 of the film is small with respect to the typical spreading lengthscale .
the dynamics of the spatiotemporal evolution of the height field @xmath1 is of a very general form , essentially a nonlinear diffusion equation .
the equation is a generalization of the so - called porous medium equation @xcite which is used to describe motion transport in soils .
one interesting problem is that of ` focusing flows ' , describing the advancing flow of a surrounding fluid to flood an interior dry spot . the simplest case would be of an axisymmetric fluid layer @xmath2 impinging on a circular hole .
solutions for the porous medium equation were shown to have self - similar structure @xcite .
self - similar solutions are described by the form @xmath3 in this representation , @xmath0 is decomposed into @xmath4 , describing the time - dependent scaling of the height profile , an @xmath5 spatial profile function and @xmath6 giving the scaling of the radial variable .
self - similar solutions occur in many contexts @xcite .
when the scaling functions in ( [ eq : selfsim ] ) can be obtained analytically from dimensional analysis and physical constraints as power laws with rational exponents , ( [ eq : selfsim ] ) is called a first - kind similarity solution . however , this approach applied to the problem for converging flows does not uniquely determine values for the exponents .
being a second - kind similarity solution @xcite , the scaling exponent must be obtained as part of the solution of a differential equation problem for the similarity profile @xmath5 . for convergent
viscous gravity currents , in which the main driving force is gravitational , the radius of the dry spot was predicted to collapse like @xmath7 with @xmath8 where @xmath9 is the point of collapse @xcite .
+ at smaller length scales , _ surface tension _ may also play a role in the dynamics of gravity currents .
much like gravity , surface tension also has a similar tendency to smooth out height variations in @xmath2 .
the capillary length @xmath10 , the ratio between surface tension forces and gravity , set by surface tension @xmath11 , and @xmath12 respectively , gives a typical lengthscale on which surface tension effects are significant .
under such conditions , surface tension gradients can dominate the spreading dynamics of the flow @xcite .
the role of surface tension forces in thin films with considerable gravitational driving has received some attention @xcite . the methods to estimate
scaling exponents that we introduce here will allow us to map out collapse behavior for fluids driven by an arbitrary ratio of gravitational and capillary forces .
we study converging viscous gravity currents with surface tension through experiments and numerics .
experimentally , we create an axisymmetric thin film by fast spin - up of a fluid in a confined container . stopping the rotation
subsequently yields a focusing flow of the fluid to re - fill the central region .
we obtain numerical results for comparison by solving the time dependent axisymmetric lubrication equation for the same system . to compare analytic results to experiments or numerical results
, one must extract such scaling exponents from either experimental or numerical surface structure @xmath2 data .
this is often a huge challenge due to the noise on @xmath2 or the finite accuracy of numerical solutions .
obtaining power law exponents is generally difficult @xcite . in the case of self - similar forms
, the problem is compounded through the fact that multiple exponents have to be obtained simultaneously .
we present here two practical methods for obtaining such exponents from experimental or numerical data .
we apply the methods to probe convergent fluid films which are driven by both gravity and surface tension .
the presence of surface tension in the driving of the collapse is expected to cause deviations from the diez scaling .
we indeed observe such deviations by applying our methods over the entire range from purely gravity to purely surface tension dominated flows .
we find good agreement between the methods , and find a reasonable match between numerically obtained data and experimental results .
the scaling form of eq .
[ eq : selfsim ] can in fact even be used in the regime where both driving forces play a role in the fluid flow .
our methods are of general interest in contexts where scaling forms are to be expected and wherever experimental or numerical data can be cast in scaling form .
the article is arranged as follows : first , we discuss the practical background of converging fluid films driven by both gravity and surface tension .
then we discuss in sec .
[ sec : exp ] the experimental set - up used to obtain experimental data , including the setting up of initial conditions for the collapse .
this also provides the context for the type of numerical analysis we will do . in sec .
[ sec : num ] we introduce the theoretical background , the relevant scaling parameters and the numerical methods and we elaborate on the numerical scheme used to obtain solutions to the thin film equation .
the numerical results reveal the typical difficulty of extracting exponents with log - fitting of power laws , both in experimental and numerical data .
we then describe the two methods to extract scaling exponents , using our numerical data as a guide
. the effectiveness of the methods is then demonstrated , and used to explore the transition in dynamics from gravity driven , to surface tension driven flows .
we also compare the numerical results with some experimental data .
( color online ) ( a ) schematic drawing of the container and all the relevant parameters : @xmath13 the rotation speed , @xmath14 the initial filling height , @xmath15 the radius of the container , @xmath16 the viscosity and surface tension of the fluid respectively .
( b ) sketch of the initial conditions for the collapse dynamics in experiments ( upper figure ) .
these are somewhat different from the initial conditions used in the numerics ( lower figure ) .
( c ) schematic drawing of the interferometry setup .
( d ) schematic of the surface structure scanning experiment .
( e ) a typical interferometry image from a dry spot .
( f ) ctf size versus rotation rate : experimental observations for @xmath17 mm , numerical results ( solid lines , see text ) and the mass conservation constraint @xcite ( dash - dot ) for comparison.,width=321 ] the experimental approach to study these convergent flows starts by first creating an axisymmetric reservoir of fluid in a rotating container .
fluid confinement inside the container leads trivially to the following heuristic picture : during rotation , centrifugal forces drive the fluid to the outer edge of the container . cessation of the rotation then ` releases ' the fluid and creates an axisymmetric flow .
there are many experimental advantages of this approach : it creates highly reproducible initial conditions in a mechanically simple and small set - up .
the dependence on experimental parameters such as rotation rate , fluid viscosity and precursor film thickness can be easily tested .
there is full optical access to the complete fluid film height profile , via various techniques .
determining the collapsing surface structure is critical to determining the collapse dynamics , especially the possibly self - similar structure of the fluid film close to the time of collapse .
+ the experimental system consists of an initially @xmath181 mm thick layer of fluid in a shallow cylindrical container see fig .
[ fig : setup]a .
the container is rotated using a stepper motor with closed loop controller ( parker zeta drive 6104 ) .
the closed loop stepper motor controller can be programmed to run any time dependent rotation speed profile @xmath19 with a maximum of two rotations per second ( rps ) .
the container measures 13 cm in diameter and 2 cm in height .
to fix the temperature dependent viscosity @xmath20 and surface tension @xmath11 of the fluid , the container is uniformly heated to a temperature of 24@xmath21c unless otherwise noted by running water at a set temperature through the double - walled rotating axis for details see @xcite . on the base of the container , a silicon wafer ( university wafers )
is placed .
the wafer is fixed to the base through the deposition of a small ( @xmath22 ml ) amount of fluid between wafer and the container .
suction force remains even after complete submersion of the wafer .
the container is filled with a volume @xmath23 of fluid which gives an initial filling height @xmath24 with @xmath15 being the radius of the container .
we use polydimethylsiloxane ( pdms ) for all experiments described in this work ; this fluid completely wets the silicon wafer .
the transparency of the pdms and reflectivity of the silicon wafer allows for a laser - assisted alignment of the gravity - leveled fluid surface and the silicon wafer in the container , whose orientation can be tuned by set screws .
interferometry provides access to the spatial structure in the thin film dynamics
see fig [ fig : setup]c , e .
container illumination is provided with a uniform sodium light via a beam splitter .
the spatial structure of the interference pattern of reflected and incoming light waves is recorded with a high resolution digital camera . + in a rotating container , a fluid will set up a parabolic surface profile to balance gravitational pressure and centrifugal forces .
however , for a shallow fluid layer , above a critical rotation rate , @xmath25 @xcite , the solid base precludes the parabolic solution and results in a piecewise - defined height profile with a central ` spot ' .
outside of the spot , for @xmath26 , the approximate profile remains parabolic . within the spot ,
the nature of the solution depends on the wetting properties of the container s base ( the silicon wafer ) . for fluids that wet the base ,
the spot will be covered by a thin film that we call the central thin film ( ctf ) . for non - wetting fluids
, the parabolic profiles will end at a sharp contact line at @xmath27 . in this paper
, we discuss only the case of completely wetting fluids , which sets up a height profile as sketched in fig .
[ fig : setup]b .
+ to find the initial conditions before the collapse created by the rotation , we then solve the steady state lubrication equation to obtain the complete surface profile @xmath28 for different rotation rates and initial volumes set by @xmath14 ; the ctf radii extracted from the numerics are in good agreement with the approximation from linden @xcite that assumes a piece - wise continuous surface structure , consisting of only a flat layer and a parabolic surface .
for such a steady state solution , mass conservation implies that @xmath29^{1/2}~,\ ] ] in which the ctf mass and surface tension at the boundary of the container @xcite are neglected .
the size of the central spot @xmath30 depends on the square root of the rotation rate @xcite and the total volume of fluid in the container .
+ we verify eq .
[ eqn : parabol ] by using interferometry to characterize the thin fluid film in the center of the container .
[ fig : setup]e shows a typical interferometric image of the ctf
. there are several features in this image . when the container is properly leveled , the ctf is axisymmetric and flat with only modest height variations at best @xcite .
the fine structure of the edge of the ctf is not visible through interferometry , so we arbitrarily but consistently define the edge of the ctf by the faint ring indicated by the arrow in fig . [
fig : setup]e , which is visible in all experiments .
the obvious fine structure within the ring is related to the contact line between the thin film and the parabolic solution .
this fine structure is nontrivial ; it is left for future work .
the radius @xmath30 of the ctf for various rotation rates and fluid volumes is shown in fig .
[ fig : setup]f .
@xmath31 above @xmath32 is as predicted by @xcite .
theoretical and numerical predictions are made with the same @xmath33 that produced the best fit to experimental data and produce reasonable agreement .
[ fig : setup]f shows also that the two theoretical methods of determining the ctf size versus @xmath13 are consistent with each other .
+ during the process of establishing a parabolic profile , the ctf is also continuously draining fluid , making the ctf change in thickness .
this drainage process is governed by a balance between centrifugal forces and viscous drag in the thinning layer .
the efflux of fluid , radially outward from the ctf , slows progressively over time towards an equilibrium height profile , as described by the scaling laws first given by emslie , bonner and peck ( ebp ) @xcite . in order to observe the collapse dynamics , it is however important to make sure that the ctf is as thin as possible . therefore , in the numerics , we set up initial conditions as shown in fig . [
fig : setup]b , where the total volume of the triangular surface area is equal to one , which can then be rescaled with the height factor @xmath14 .
we choose the central layer thickness @xmath34 sufficiently small that it does not affect the results .
+ ( color online ) ( a ) interferometric top view of a collapsing dry spot ; the circular shape outlines the edge of the collapsing dry spot .
the intensity profile of the dashed line is shown as a function of time in the space - time plot of panel ( b ) , contrast enhanced .
the spot size is indicated with the arrow ; time progresses towards the right .
total time of collapse @xmath35 150 seconds .
( c ) the spot size as a function of the time to collapse ( time increasing to the left ) for a typical @xmath36 mm , @xmath37 mpa@xmath38 after spinning for several minutes at @xmath39 rps .
blue lines for reference are power laws of exponents 0.5 and 0.6.,width=321 ] [ sec : collapse ] the fluid volume under the parabolic profile at the edge of the container serves as a reservoir that , upon cessation of the container s rotation , is ` released ' in a dam break fashion .
this creates a fluid flux towards the center of the container .
an example of this behavior is indicated in fig .
[ fig : topviewcollapse]a .
we use the following experimental procedure : starting from an initially flat fluid surface in a static container , the rotation rate is ramped up to @xmath40 , which creates the partial parabolic profile . after rotating at constant @xmath13 for a finite time
, rotation is stopped virtually instantaneously .
this removes the centrifugal force and initiates the flow of the fluid from the reservoir at the boundary .
tracer particle tracking on the surface of the fluid showed that all experiments are at low enough reynolds number @xmath41 such that rotational flow never persists more than a small fraction of the initial collapse .
the ctf spot in the center of the container thus disappears relatively slowly .
we denote the size of the dynamically changing spot with @xmath42 to distinguish it from the steady state spot size @xmath43 .
this process is indicated in fig .
[ fig : topviewcollapse]b ; these experiments are similar as the one described in ref .
note that the collapse dynamics of the front over a too - thick ctf layer ends with a standard diffusive @xmath44 exponent for a fixed height threshold .
this is due to the fact that with a thick ctf layer , any cavity is simply a dip in the fluid surface with an approximately parabolic shape ; this dip shifts its vertical position while the fluid is moving to the center . to avoid having this effect obscure the nontrivial scaling , it is essential to make the ctf as thin as possible just before the collapse .
the container is spun sufficiently long to ensure this is the case .
imaging suggests that the typical initial film thickness is about 100 micrometer or less .
+ at the end of the collapse , just before cavity closure , the spot size versus time is clearly nonlinear : from the data shown in fig .
[ fig : topviewcollapse]b , the size of the shrinking ctf spot is tracked with an intensity threshold technique .
the result is shown in fig .
[ fig : topviewcollapse]c : the spot size varies as @xmath45 with an uncertainty of about 0.05 in the exponent .
however , as we will see below , a log - fitting routine is unable to capture an accurate exponent for the collapse . to better characterize the collapse dynamics
, we turn to surface structure imaging .
( color online ) ( a ) typical image ( cropped ) obtained by the camera . from a sequence of such images
we can reconstruct the full height profile @xmath2 as shown in color , indicating the height .
( b ) @xmath46 for different @xmath47 .
@xmath48 refers to the rotational forcing stage which sets up the initial parabolic profile , indicated in blue , from @xcite .
( c ) @xmath49 for different @xmath50 as indicated .
@xmath51 for this experiment .
rotation is stopped just before @xmath52 sec .
( d ) as in ( b ) measured for the same fluid at two different temperatures .
time is rescaled with the temperature dependent viscosity ( @xmath53 300 mpa@xmath54 36 mpa@xmath38 ) to indicate the similarity of the profiles.,width=321 ] to obtain full surface structure data in the experiments , we step away from interferometry and instead image a cross section of the ( radially symmetric ) height profile with a high speed camera ( photron ) see fig .
[ fig : collapse ] .
surface contrast is created by adding an oil - soluble fluorescent dye in the fluid ( pyrromethene 567 ) and illuminating the dye with a 532 nm laser line .
there is sufficient dye concentration in the pdms such that only the fluid close to the surface fluoresces ( fig .
[ fig : collapse]a ) .
+ the laser line scanning method is used to obtain the data in fig .
[ fig : collapse]a - d : here a 2.9 mm layer of @xmath18 700 mpas pdms was spun for a minute at @xmath39 rps before cessation of rotation at @xmath55 seconds . the complete structure and typical behavior of the thin film surface
is shown in fig .
[ fig : collapse]a , much like ref .
the radial surface profiles , @xmath2 , for e.g. @xmath56 and @xmath57 seconds are shown in fig .
[ fig : collapse]b . at @xmath48
there is a parabolic surface profile due to centrifugal force , which is quantitatively in agreement with the prediction from @xcite .
the equation describing the surface profile at @xmath40 : @xmath58 , with @xmath59 . for @xmath60 ,
we expect @xmath61 , which is below the vertical resolution shown in fig .
[ fig : collapse]b . after the collapse , the parabolic surface profile first inverts its curvature to form the moving front shape as visible in fig . [
fig : collapse]b at 1.4 seconds . called waiting - time behavior
, the change in curvature happens while the contact line is essentially static @xcite . beyond this point a symmetric collapse of the central cavity is observed . in fig .
[ fig : collapse]c , @xmath49 for @xmath62 and @xmath63 mm radial positions along the diameter marked by the laser line in the @xmath64 mm radius container is shown .
contour lines for both positive and negative @xmath50 are shown to indicate the radial symmetry in the surface shape ; they overlap almost entirely . as shown in fig .
[ fig : collapse]d , the time - dependence of the collapse dynamics scales inversely with the viscosity .
+ to extract the ctf collapse exponent , we need to measure both the size @xmath42 and the collapse time @xmath9 .
we do this by choosing a single height threshold : we define the radius of the ctf with @xmath65 the largest @xmath50 for which the experimentally determined height profile @xmath2 is smaller than threshold @xmath66 .
the height threshold also sets @xmath9 with some uncertainty .
[ fig : collapse]c shows that the collapse time is not easy to precisely pin - point from experimental data .
the precision in determining @xmath9 is set by the slope of time dependence for the height at the origin around the collapse time .
if the slope of @xmath67 around the collapse time is large , @xmath9 does not depend sensitively on the height threshold @xmath66 .
we however observe in fig .
[ fig : collapse]c that the slope of @xmath68 is finite , so every @xmath66 threshold comes with its own estimate of @xmath9 .
+ we show @xmath42 for two different collapse experiments and four different @xmath66 for each experiment in fig .
[ fig : expcollapse]a , b . for the experimental settings used , the data in fig .
[ fig : expcollapse]b shows that @xmath69 with an uncertainty in the exponent of about 0.05 , but very similar to the exponent obtained with the top view experiments .
these two front tracking methods therefore yield consistent results .
the exponents found with these two methods are however significantly different from the @xmath70 exponent expected for gravity driven fluids , suggesting that in our experiment , surface tension plays a role . to probe the role of both driving forces
, we will look at the thin film equation in more detail in the next section .
_ choosing @xmath71 _ as shown below , the collapse time strongly decreases with the initial layer thickness @xmath14 , and as such , it is favorable to choose thicker layers to keep experimental timescales manageable .
however , for thicker layers the precursor film needs prohibitively long spinning time to become sufficiently thin at the time of the collapse .
one solution for this challenge is to speed up the collapse by choosing less viscous pdms , but for fluid viscosities of less than @xmath18 50 mpa@xmath72s the stokes flow approximation breaks down during the collapse .
we are hence limited in choosing @xmath73 mpa@xmath72s and @xmath74 mm .
( color online ) ( a ) @xmath42 measurement for a @xmath75 mm pdms layer at 24@xmath76 , for which @xmath77 100 mpa@xmath72s . for the surface profile obtained in this experiment
we extract different @xmath42 with different height thresholds @xmath78 .
( b ) @xmath42 measurement for a @xmath75 mm pdms layer at 45@xmath21c with @xmath77 36 mpa@xmath72s . @xmath42 obtained with different height thresholds @xmath79 . in both panels , power laws of 0.55,0.65,0.75 indicated for reference .
the dash - dotted line indicates the spatial resolution of the experiment.,width=321 ]
we consider thin film dynamics in the geometry shown in fig .
[ fig : setup]a . in the low reynolds number , creeping viscous flow limit , the time dependent film height @xmath2 in the rotating container is described with a time dependent axisymmetric lubrication equation that includes surface tension , gravity , centrifugal force and disjoining pressure : @xmath80 \right .
\nonumber\\ \qquad \left .
+ \frac{\gamma}{3\eta } rh^3 \frac{\partial}{\partial r } \left [ \frac{1}{r } \frac{\partial}{\partial r}\left(r \frac{\partial h}{\partial r } \right)\right ] \right\ } = 0~. \end{aligned}\ ] ] here , @xmath2 is the axisymmetric surface height profile depending on the radial coordinate , @xmath50 , and time , @xmath47 .
viscosity is given by @xmath20 , @xmath13 is the rotation rate and the surface tension coefficient is @xmath11 .
the local pressure inside the fluid layer depends both on gravity and the disjoining pressure with the hamaker constant negative for complete wetting and of order @xmath81 j typically disjoining pressure effects are insignificant @xcite .
first , starting from a uniform film of height @xmath82 , the system is spun up with rotation rate @xmath13 .
when the profile is sufficiently close to equilibrium ( and the ctf is sufficiently thin ) , the rotation is stopped , so @xmath83 in ( [ eq : filmfull ] ) for the rest of the evolution . the concomitant simplification of eq . [ eq : filmfull ] allows for nondimensionalization .
a natural choice is to set @xmath84 and @xmath85 .
we are then left to set a length scale for the radial coordinate .
we can follow schwartz and roy in their drop spreading analysis for spin coating context @xcite and bertozzi _ et al _ @xcite and choose @xmath86 .
this yields the following equation in the absence of rotation ( @xmath83 ) , neglecting the disjoining pressure , and dropping tildes after rewriting : @xmath87 \right\ } = 0~ , \end{aligned}\ ] ] with @xmath88 the alternative choice of @xmath89 is perhaps more natural , as one would expect that the container radius is not a relevant length scale for the convergent flows in the center of the container .
the @xmath89 rescaling only modifies the prefactors : @xmath90 here the total time it takes for the collapse to complete scales with @xmath91 with @xmath9 the collapse time .
we compare this scaling with experiments introduced in sec .
[ sec : topview ] .
we measure the collapse timescale by direct imaging the ctf during collapse for a @xmath92 mpas and a range of @xmath93 to @xmath94 mm .
results are shown in fig .
[ fig : numsurf]a ; in the accessible range of data , the experimental results are consistent with a total collapse time scaling of @xmath95 , clearly incompatible with scaling from the nondimensionalization through @xmath89 .
we therefore choose @xmath86 , which does produce a @xmath95 scaling of the total collapse time .
the properties of pdms , with density @xmath96 kg / m@xmath97 and @xmath98 n / m together with the container radius @xmath99 cm fixes @xmath100 in all our experiments , with only a weak dependence on temperature through @xmath101 n / km .
+ ( color online ) ( a ) collapse time versus filling height @xmath14 ; indicated is a power law with exponent @xmath102 .
green pluses are experimental results ( see text ) .
red dots are numerical results .
the timescale in the numerics is arbitrary ; this limits us to compare only the scaling of @xmath14 with the top view experiments .
( b ) typical @xmath2 profile produced by the numerics .
color indicates height , in units of @xmath14 .
initial conditions are as shown in fig .
[ fig : setup]b.,width=321 ] we use a second - order - accurate implicit finite difference scheme to solve the time dependent axisymmetric lubrication equation ( eq . [ eq : nondimfilmfull ] ) .
time stepping is dynamic to allow us to resolve the fluid surface motion just before collapse .
numerics give us access to a wide range of radii @xmath103 m ( using @xmath104 to indicate a range ) and surface tensions @xmath105 n / m , which allow us to study a large range of @xmath106 .
the ratio of @xmath20 and the time step in the numerics is set to keep computation time small .
we ensure mass conservation inside the film to within @xmath107 or better . + a typical @xmath2 is shown in fig .
[ fig : numsurf]b .
the initial conditions , set up to mimic the surface profile during rotation , clearly evolve towards a collapsing ctf upon cessation of the rotation .
from these data , we extract several quantities .
first , we extract the collapse time scaling by measuring the time it takes for the thin film height @xmath0 at @xmath108 to rise to the substantial height of @xmath109 , which is much larger than the initial film height @xmath110 .
the results are insensitive to the choice of these parameters , but they induce an arbitrariness in the time scale through the choice of threshold and the choice of the viscosity .
we exploit this arbitrariness to shift the timescales obtained by the numerics to an order of magnitude where they overlap with experimental data .
we can still extract the dependence of the total collapse time on @xmath14 .
the numerical collapse dynamics are indeed consistent with the experimentally observed trends shown in fig .
[ fig : numsurf]a : this timescale is @xmath111 , as expected from eq .
[ eq : speed ] .
+ second , the size , @xmath42 , of the ctf or the ` spot ' in the center of the container is shown in fig .
[ fig : numlog]a , where @xmath42 is shown as a function of the time to collapse @xmath112 with @xmath9 the collapse time , defined as the moment at which all film heights exceed a minimum threshold level , @xmath113 .
+ the extracted scaling exponents tend to the value predicted by diez _ et .
_ @xcite for large values of @xmath114 .
however , the exponents extracted with log fitting show considerable scatter and no systematic convergence towards the @xmath70 predicted by diez _ et .
_ , indicated by the dash - dotted line in fig .
[ fig : numlog]b .
we also find no trend of the extracted exponents with @xmath14 or the thresholds used for the extraction of @xmath42 ( not shown ) .
this suggests that a better method to extract exponents is needed .
( a ) @xmath42 for a typical numerical at @xmath115 and for threshold @xmath116 .
( b ) exponents from log fitting over a range of @xmath114 where the exponent is expected to be @xmath70 ( dash - dotted line ) .
the error in fitting due to @xmath66-threshold variations is of the order of the symbol size .
grayscale data points have a black border for visibility ( as in other figures).,width=321 ]
to extract scaling exponents for the collapse dynamics , we can make use of the special properties of self - similar functions .
our two exponent extraction methods rely on assuming that the thin film equation eq .
[ eq : filmfull ] allows for similarity solutions of the form @xmath117 in which @xmath118 are two scaling exponents , @xmath119 is a shape function and @xmath9 is the collapse time . in the notation of eq .
[ eq : selfsim ] , we thus set @xmath120 and @xmath121 , which means that the dry spot with diminishing size @xmath122 .
we can use this feature to find the @xmath118 for a particular experimental or numerical solution of the @xmath2 profile , given physical variables such as surface tension , density , et cetera .
in particular , we can find constraints on @xmath123 . plugging ( [ eq : scalingform ] ) into ( [ eq : nondimfilmfull ] )
, we find that in the large @xmath114 limit , a relation @xmath124 is required to give self similar behavior and an ordinary differential equation ( ode ) for the shape function @xmath119 : @xmath125 here @xmath126 is the reduced variable and primes denote @xmath127 , differentiation with respect to the reduced variable . in the small @xmath114 limit , we can do the same substitution for @xmath2 . in this regime , we find @xmath128 is needed to do the same separation of the shape function dynamics from the time - dependent scaling prefactor .
we are thus left to find @xmath129 and @xmath130 over the entire range of @xmath114 . as a function of band height , extracted with the least squares method described in the text .
grayscale indicating the @xmath131 value of the nondimensionalized least squares error is shown on the right .
( c ) band height for which the best @xmath118 solution was found , as a function of @xmath114 .
grayscale is the same as in ( a , b ) .
( d ) the relation @xmath123 found with the least squares method ; the dash - dotted line shows @xmath132 and the dashed line indicates @xmath133 , the limiting behaviors predicted for high and low @xmath114 ( indicated by gray level ) respectively .
[ fig : ls_app],width=321 ] the first method to generally extract scaling exponents from numerical or experimental data relies on knowledge of the spatial and temporal derivatives @xmath134 respectively . for the solution of the scaling form of eq .
[ eq : scalingform ] , some algebra provides an expression relating @xmath129 and @xmath130 : @xmath135 this relation should be valid for all @xmath114 and at each @xmath136 if the solution is self - similar . having access to the gradients of @xmath0 , especially in the numerics , eq . [ eq : lsform ] provides a set of linearly independent relations , with only @xmath129 and @xmath130 as fit parameters
. we can thus determine @xmath118 to solve eq .
[ eq : lsform ] in a least squares sense .
[ eq : lsform ] should be valid for all @xmath2 near collapse ; however , we can further maximize the resolving power of this exponent extraction method by applying this relationship to _ height bands _ within the surface .
the band selection @xmath137 selects a band of data points with a thickness of @xmath138 .
we choose these bands anywhere between @xmath139 to @xmath140 . for each numerical data set in a particular band
we compute an @xmath118 with ( [ eq : lsform ] ) .
we always use the data for which @xmath141 and limit ourselves to a time period close to the collapse , of five percent of the total collapse duration : @xmath142 .
the results for a typical run are shown in fig .
[ fig : ls_app]a , b for @xmath130 and @xmath129 respectively . in these panels ,
the color coding indicates the least squares error @xmath143 of all data in one height band .
both figures show similar trends : for low and high bands close to the container base and to the surface respectively , the @xmath118 found with the least squares method give poor fits , as indicated by the large error . in an intermediate regime
the error has a clear minimum .
this minimum is consistently of order @xmath144 , as shown in fig .
[ fig : ls_app]c .
this panel also indicates at which band the best fit is found ; the optimal band is around @xmath145 for large @xmath114 , moves up in the intermediate regime and is lower for small @xmath114 .
we verify that the least - squares method gives the relation between @xmath129 and @xmath130 in the large and small @xmath114 limit in fig [ fig : ls_app]d , where we indicate @xmath123 and two limiting behavior in the dashed and dash - dotted line .
interestingly , all data is to be found on either one of the functions from eqs .
[ eq : alphagrav ] and [ eq : alphacap ] in the limiting cases , with the transition occurring around @xmath146 .
+ obtained for @xmath147 n / m and @xmath148 m ; @xmath149 .
as an example of a combination of @xmath118 that give a poor collapse , here we take @xmath150 .
the light gray lines represent the triangulation of the data points .
grayscale indicates the nondimensionalized @xmath151 .
( b ) the probability distribution function of the surface areas within the triangulation from ( a ) .
the dash - dotted line indicates the median @xmath152 of the set @xmath153 .
( c ) @xmath154 for this particular numerical run .
grayscale indicates ranges from low ( black ) to high ( white ) .
( d ) same as ( a ) , without triangulations and with @xmath155 , representing the collapse properties indicated by the @xmath156 in ( c ) , the @xmath118 for which @xmath152 is minimal .
note that we suppress dimensions in the various logarithmic axes.[fig : vtrimethod ] , width=321 ] the least squares method is very useful in finding scaling exponents for numerical data .
however , in experimental data , obtaining derivatives @xmath157 is often challenging as derivatives obtained from experimental data are generally noisy .
indeed , we found that using the least squares method on experimental data does not unambiguously pick an @xmath118 pair independently of the thresholds required in the analysis .
it would therefore be advantageous to have another method to extract scaling exponents .
+ our second method works by plotting all individual data points obtained for a measured surface profile @xmath2 on a rescaled cartesian coordinate axis @xmath158 . in general , this plotting method will produce a scattered set of data points .
the crux of this method is that for the @xmath118 that best represents the scaling function , the aforementioned scatter plot of all the surface points collapses to produce a _ curve _ in the aforementioned coordinate system .
this curve represents the function @xmath119 from eq .
[ eq : scalingform ] . for data points on this line
, the delaunay triangulation @xcite of the set will produce mostly triangles with a very small area , and only a few large triangles to complete the convex hull of the entire data set .
however , for a choice of @xmath118 that does not represent the scaling behavior , the delaunay triangulation will produce a much larger number of large triangles . by optimizing the number of small triangles in the delaunay triangulation of the data points on the scatter plot
, we can thus find the @xmath118 pair that produces the best collapse of the data . in the delaunay triangulation
, every triangle covers an area @xmath23 .
the entire dataset makes a set of triangulated areas @xmath159 .
an example of such a set and its triangulation is shown in fig .
[ fig : vtrimethod]a for _ numerical _ results and a reasonable , but not best choice of @xmath118 . to maximize the number of small triangles
, we look at the statistical properties of the set @xmath159 . the probability distribution function for finding a triangle of area @xmath23
is shown in fig .
[ fig : vtrimethod]b .
this panel shows the distribution of areas found for the example from fig . [
fig : vtrimethod]a .
we find that the median @xmath152 of the set @xmath159 captures the number of small triangles well : a smaller median indicates more small triangles are present in the set . in order to find the best collapse
, we thus have to compute @xmath154 and find its minimum . in fig .
[ fig : vtrimethod]c , we show @xmath154 .c , although its effect is small . ]
the minimum of the triangulated area is indicated by the cross , whose collapse is shown in fig . [
fig : vtrimethod]d : indeed the minimum in @xmath154 corresponds to a very good collapse of the data .
[ fig : vtrimethod]d therefore also shows the form of @xmath119 .
note that in this example , we again extract all @xmath2 for which @xmath141 and @xmath160 , meaning all @xmath47 before the collapse , but sufficiently close to the collapse that the scaling assumptions should hold .
+ versus @xmath114 as determined by the triangulation method .
grayscale indicates @xmath129 .
the error bar indicates the uncertainty in the exponent due to the time range threshold set in the analysis .
the dash - dotted line indicates the @xmath70 exponent expected in the large @xmath114 limit .
( b ) @xmath123 with data from ( a ) ; grayscale @xmath161 .
the dashed and dash - dotted lines indicates the constraint on @xmath123 , which are clearly satisfied in the large and small @xmath114 limit.[fig : vtriexponents],width=321 ] to verify the triangulation method over a range of @xmath114 , we apply the method to the series of numerical simulations on which we also tested the least squares methods .
we extract @xmath118 for each of the simulations .
results are shown in fig .
[ fig : vtriexponents]a , b . fig .
[ fig : vtriexponents]a shows @xmath130 versus the dimensionless ratio @xmath162 ( the grayscale indicates the value of @xmath129 ) .
this panel shows that there are two regimes : for large @xmath114 , the @xmath130 that best describe the shape of @xmath2 asymptotes to the expected value of @xmath70 indicated with the dashed - dotted line . for smaller @xmath114
there is a large crossover range to a regime in which @xmath163 . in fig .
[ fig : vtriexponents]b we show @xmath123 . scaling theory predicts the relationship between the exponents @xmath123 as indicated by the two lines .
the triangulation method finds the large and small @xmath114 limiting behavior ; at intermediate @xmath114 , the method produces @xmath118 that deviate from the expected constraints .
+ the numerical test of the triangulation method showed that a minimum triangle area median can unambiguously be defined , which picks a unique @xmath118 pair .
we can now use the triangulation method to extract scaling exponents in experimental data . for a fluid collapse experiment run at 45 @xmath76 , the quantity @xmath154 for @xmath164
is shown in fig .
[ fig : vtriexp]a .
the overall structure of the experimentally determined @xmath154 is very similar to the numerical one ; for comparison , see fig .
[ fig : vtrimethod]d .
the triangulation method is sensitive to the total amount of surface structure data used in the triangulation of @xmath119 , and for the resolution at which we test all pairs of @xmath165 to work .
this sensitivity is quantified in fig .
[ fig : vtriexp]b .
we show there @xmath154 for several time ranges @xmath166 and three different resolutions at which we looked for minima in @xmath118 space .
for @xmath167 , the @xmath118 pairs found approach the low @xmath114 limit function @xmath133 .
however , if we include more data before the collapse point into account to estimate the minimum median triangulation area , we find that @xmath118 quickly deviate from this low @xmath114 limit .
the best collapse data should be available just before the collapse , so we conservatively estimate @xmath168 and @xmath169 .
+ ( a ) @xmath154 for a particular time cutoff of 0.2 seconds for the 45 @xmath76 pdms data set shown in fig .
[ fig : expcollapse]b .
the grayscale is arbitrary ; black indicates a low value for @xmath154 .
( b ) minima in @xmath154 correspond to a pair @xmath118 .
pair location depends on @xmath118 scan resolution : @xmath170 , @xmath171 , @xmath172 and the percentage of the total collapse time @xmath9 included in the triangulation set ( grayscale ) .
numerical data is always obtained with @xmath173 of @xmath9 ( @xmath174 ) .
based on these thresholds , the error bars on the determination of @xmath118 are as indicated in fig .
[ fig : ls_comp]a , b.,width=321 ]
we summarize the results of the analysis , the two methods to extract scaling exponents and the experimental data in fig .
[ fig : ls_comp]a .
there are a few important observations : _
( i ) _ the least squares method very well satisfies the derived relations for @xmath123 at all @xmath114 , even though these relations are technically only valid in the limits of very small and very large @xmath114 . _
( ii ) _ the triangulation method results are in agreement with the least squares method in the large and small bond number limits . _ ( iii ) _ the single experimental data point that we have at @xmath175 , although representative for many repeated experiments , only agrees with the predicted relations for the most favorable choice of @xmath9-cutoff in the analysis ( see fig .
[ fig : vtriexp]b ) . + especially point _ ( iii ) _ warrants a discussion . in the triangulation method
, we can only get full agreement between experiment and theoretical values expected for @xmath118 for the most favorable cutoff in @xmath9 .
this can hardly be called a robust result .
however , for just the exponent @xmath130 , the consistency with numerical results is reasonable .
apart from the triangulation method , also the independent top - view experiments and simple thresholding techniques we described in sec .
[ sec : exp ] give us values for @xmath130 .
we obtained an exponent @xmath176 . with @xmath100 ,
the experiments are in an intermediate regime of bond numbers , where @xmath130 is expected to deviate from the diez result , as can be seen in fig .
[ fig : ls_comp]b .
the value obtained for @xmath130 in our experiment is thus not entirely unrealistic .
+ nevertheless , @xmath129 is not matching , even though the triangulation method is expected to work as per our calibration of the triangulation method on the numerical data .
there are two possible explanations for the deviation : either we have missed a systematic experimental bias in our methods , or the thin film equation eq .
[ eq : filmfull ] does not describe the dynamics very well close to the collapse .
one approximation in the thin film equation is that surface gradients are small @xmath177 .
close to the collapse moment , the fluid surface however forms an axisymmetric `` funnel '' touching the base of the container , in which this approximation seems less valid .
in particular the strength of the loop curvature contribution will become relevant close to the collapse .
the strength of this term in the laplacian of the surface shape also depends on the smallness of the surface gradient terms , which in the current expression may not be fully appreciated .
+ the bond number that we used to quantify the transition from gravity to surface tension driven dynamics does not allow one to experimentally probe a large range , as the material parameters included are difficult to vary over many orders of magnitude .
we chose this bond number definition to make the collapse timescale dependence on the filling height @xmath14 consistent with experiments .
it can however be argued that just before the collapse the dynamics is indeed not anymore dependent on boundary conditions such as the total volume ( and hence @xmath15 ) and that a different length scale should be chosen to nondimensionalize the thin film equations .
for example , one could let the bond number scale with @xmath178 instead of @xmath179 . in these thin film applications typically @xmath180 , which would push @xmath114 at least four orders of magnitude lower , and the transition point between gravity driven and surface tension driven flows to @xmath181 , a more reasonable number .
given the mentioned uncertainties of the curvature approximations already included in the current modeling , we leave these questions for future work . for both least squares method ( @xmath182 ) and the triangulation method ( @xmath156 ) .
grayscale indicates @xmath183 ; the dash - dotted line shows @xmath132 and the dashed line indicates @xmath133 .
the experimental data point in ( a , b ) comes from analysis discussed in sec .
[ sec : tri ] ; the error bars are conservative .
( b ) @xmath184 for both least squares ( filled circles ) and the triangulation method ( @xmath156 ) .
the asymptotic 0.762 limit is indicated with the dash - dotted line .
grayscale indicates @xmath185 used in each numerical datapoint ( filled circles only ) .
the black datapoint again indicates the experimental value obtained from triangulation analysis.[fig : ls_comp],width=321 ]
we studied the effect of surface tension on viscous gravity currents in a novel experimental setup , quantifying the relative strength of the surface tension with a bond number that depends on the reservoir radius .
we show that the converging surface profile is of self - similar form in the surface tension dominated regime , and we derive relations between exponents for both bo limits .
extracting power - law exponents on log - log scale turns out to not be accurate enough in the context of the exponents of similarity solutions .
we thus develop two complementary methods to extract scaling information from the self - similar structure of the converging surface profile .
these methods allow us to analyze numerical data over the complete range of purely gravity driven , to purely surface tension dominated flows , and allow for a comparison with experimental data .
our experimental data seems to suggest that the collapse regime is perhaps not well described by the thin film approximation .
however , the exponent extraction methods developed work very well for the numerical data and give results in excellent agreement with the theory .
_ acknowledgements _ we thank joshua bostwick , roberto camassa , rich mclaughlin , omar matar and howard stone for stimulating discussions , and the reviewers for their patience and valuable feedback .
this project was funded by nsf dms0968252 . | the surface structure of converging thin fluid films displays self - similar behavior , as was shown in the work by diez et al [ q. appl .
math 210 , 155 , 1990 ] . extracting
the related similarity scaling exponents from either numerical or experimental data is non - trivial . here
we provide two such methods .
we apply them to experimental and numerical data on converging fluid films driven by both surface tension and gravitational forcing . in the limit of pure gravitational driving
, we recover diez semi - analytic result , but our methods also allow us to explore the entire regime of mixed capillary and gravitational driving , up to entirely surface tension driven flows .
we find scaling forms of smoothly varying exponents up to surprisingly small bond numbers .
our experimental results are in reasonable agreement with our numerical simulations , which confirm theoretically obtained relations between the scaling exponents . |
h ii galaxies are dwarf emission - line galaxies undergoing a burst of star formation .
they are characterized by strong and narrow emission lines originated in a giant star - forming region which dominate their observable properties at optical wavelengths ( e.g. @xcite ) .
most are blue compact galaxies ( bcgs ) .
they have very low metallicities , high rates of star formation and a very young stellar content . many are compact and isolated .
one of the reasons why these objects have attracted significant attention is the possibility that they are very young galaxies in the process of formation .
this possibility , however , has been challenged since evidence for an old ( several gyr ) stellar population has been found in numerous bcgs ( see @xcite for a review ) .
therefore , a model with a single , instantaneous burst of star formation does not seem appropriate to describe these galaxies and a succession of short starbursts separated by quiescent periods seems more likely ( @xcite ) . even in the case
if hii galaxies are not primeval , little chemical evolution has happened and they provide important information about how galaxies form and evolve , as well as about the process of star formation in low metallicity environments . the most distant hii galaxy in the catalogue compiled by terlevich et al .
( 1991 ) is at redshift @xmath190.31 , while the vast majority are at @xmath200.1 .
the proximity of these galaxies allows studies of their structure , metal content and stellar population with high sensitivity and precision , difficult to achieve for high redshift star forming galaxies .
still , finding high redshift hii galaxies is of great interest to investigate their abundance patterns and stellar properties at earlier epochs , compare them with the nearby counterparts and obtain further information about the formation and evolution of galaxies using objects that have undergone little chemical evolution .
the lynx arc is a star forming object at [email protected] .
it was discovered during spectroscopic follow - up of the cluster rxj 0848 + 4456 ( z=0.57 ) from the @xmath22 deep cluster survey ( @xcite ) .
the arc shows a very red r - k color and very strong optical and uv ( rest frame ) narrow emission lines . from the analysis of hst wfpc2 images and
keck optical and near infrared spectroscopy , fosbury et al .
( 2003 , fosb03 hereafter ) concluded that the arc is an hii galaxy magnified by a factor of @xmath910 by a complex intervening cluster environment .
the authors concluded that the continuum is mostly nebular . by means of photoionization modeling
, @xcite showed that the spectroscopic properties of the lynx arc are all consistent with a simple hii region model of a cluster of @xmath910@xmath23 massive stars , characterized by a very high t@xmath2480 000 k , a high ionizing parameter ( [email protected] ) and low nebular metallicity z@xmath9 5% z@xmath11 .
indirect arguments suggest that the ionizing stellar population could have much lower abundances . using simple photoionizaton models ,
the authors conclude that the spectroscopic properties of the lynx arc are consistent with those of a metal - poor nebula ionized by a cluster of primordial ( i.e. metal - free ) stars . in this scenario ,
the overabundance of si implied by the models is explained as due to enrichment by past pair - instability supernovae , requiring stars more massive than 120 m@xmath11 .
we investigate in this paper the possibility that normal ( i.e. non primordial ) stars are responsible for the excitation of the gas in the lynx arc .
our first goal is to constrain the physical properties of the gas : electron temperature , density , metallicity .
we will first use standard spectroscopic techniques of nebular analysis ( e.g. @xcite ) .
these results will be compared with those from detailed photoionization modeling and those obtained by @xcite .
the second goal is to characterize the ionizing stellar population : age , metallicity , stellar mass of the burst , possible presence of wr stars .
this will be achieved by means of detailed photoionization models , whose objective is to find an ionizing stellar population able to explain all the spectroscopic properties of the lynx arc ( line ratios , equivalent widths , intensity of the continuum ) .
in addition , such models set tight constrains on the nebular properties , which must be consistent with the results from the standard techniques mentioned above .
we present in 2 the results of the nebular analysis and the photoionization models .
these results will be discussed in 3 .
summary and conclusions are presented in 4 .
in this section we will set constrains on the electron temperature , density and the nebular abundances of the ionized gas in the lynx arc .
all the analysis and discussion presented below are based on spectra obtained with different slit widths and orientations ( see @xcite for a detailed description of the data set and reduction techniques ) .
for this reason , @xcite estimated carefully scaling factors that were applied in order to derive a consistent calibration .
we will therefore assume that the errors introduced by calibration uncertainties are negligible .
the results are also dependent on whether reddening is present or not .
we have ignored this effect .
the very strong uv lines and the lack of evidence for dust reddening from the photoionization models and the fit to the continuum shape ( @xcite ) makes as confident that such assumption is reasonable and our results are not seriously affected by it .
our own work presented in this paper will show that the line ratios are consistent with no dust reddening .
the [ oiii]4363/[oiii]5007 ratio has been traditionally used to measure electron temperatures in ionized nebulae .
unfortunately , the [ oiii]@xmath64363 line is outside the observed spectral range .
we have used the uv [ oiii ] doublet instead , since the [ oiii]@xmath51661,1666/[oiii]@xmath65007 ratio is mostly sensitive to electron temperature .
we show in fig.1 the dependence of the [ oiii]@xmath51661,1666/[oiii]@xmath65007 ratio with electron temperature the position of the lynx arc ( the value of the ratio is [email protected] , see table 1 ) is also shown as a filled circle .
it implies @xmath2517300@xmath8 k. this value is somewhat higher than that predicted by the photoionization models presented in @xmath262.2 which give @xmath2716200@xmath10500 k , although taking errors into account , the discrepancy is not important .
this further supports that dust reddening effects are negligible , otherwise , the temperature derived from fig .
1 would be a lower limit and large discrepancies with the photoionization model predictions would arise .
this temperature is much lower than that derived by @xcite and @xcite by means of photoionization modeling ( @xmath920000 k ) .
the discrepancy is due to a combination of the very hard ionizing continuum ( very energetic electrons are therefore released in the ionization processes ) and high @xmath28 ( ionization parameter ) value used by these authors .
there are three uv line doublets we can use to estimate the electron density in the lynx arc : ciii]@xmath51907,1909 , siiii]@xmath51883,1892 and niv@xmath51483,1487 ( see @xcite , @xcite ) .
the results shown in table 1 correspond to t@xmath2910000 - 20000 k range , where the ratios are quite insensitive to temperature .
the upper limits take the errors into account ( i.e. the densities shown correspond to the minimum possible value of the line ratio ) .
the siiii ] doublet density determination has to be considered carefully .
we have noticed that the predictions by keenan , feibleman & berrington ( 1992 , see fig.2 ) are actually inconsistent with the results of our photoionization models ( 2.3 ) . as an example , for some of the @xmath30=1000 @xmath3 models presented in 2.3 , siiii]1883/1892=1.8 , which is above the maximum possible value ( @xmath91.5 ) predicted by those authors for this ratio . for models with @xmath30=10 @xmath3 , we find that the ratio reaches a value of 2 in some cases .
we also find that the ratio varies for a fixed density ( between 1.45 and 1.8 for @xmath30=1000 @xmath3 ) , depending on other parameters such as the spectral energy distribution ( sed ) properties .
.electron densities measured for the lynx arc using the ciii ] , siiii ] and niv ] uv doublets . [ cols= " < , < ,
< , < , < , < " , ] we have then calculated the stellar continuum at @xmath91600 ( rest frame ) emitted by these instantaneous bursts and compared it with the continuum level measured for the lynx arc which is @xmath90.15 @xmath31jy ( observed frame value corrected for magnification , @xcite ) . this corresponds to f@xmath32=7.89@xmath110@xmath33 erg s@xmath34 @xmath35 hz@xmath34 ( rest frame ) .
the results are shown in table 7 .
the theoretical values are a factor of @xmath910 higher than the value measured for the lynx arc .
the discrepancy is even higher if we take the nebular continuum into account , which according to @xcite is the dominant continuum component .
as @xcite , we therefore conclude that the stellar burst that reproduces the measured h@xmath18 luminosity would produce an observable continuum much ( @xmath910 times ) brighter than that detected from the lynx arc . adding the nebular contribution
, the continuum should be even brighter .
the conclusion is not affected by uncertainties on the magnification factor @xmath31 if the continuum and the h@xmath18 flux scale in the same way with @xmath31 .
a similar problem was reported by @xcite regarding the giant ly@xmath36 nebulae associated with lyman break galaxies .
the authors found that the measured uv continuum is not enough to explain the high ly@xmath36 luminosities of the nebulae .
we now discuss some possible reasons for this discrepancy in the lynx arc .
_ the agn scenario _
the previous discussion suggests that the ionizing source in the lynx arc is hidden from the observers view .
such a geometry is characteristic of type ii active galaxies ( agn ) .
@xcite have recently suggested that the lyman `` blobs '' found by @xcite might be powered by a hidden agn .
@xcite suggested the possibility that the lynx arc is a type ii active galaxy .
we , however , strongly support that this is a star forming object for several reasons : * an unusual ( filtered ) ionizing continuum shape is required to explain the line ratios of the arc ( binette et al .
2003 ) within the agn scenario .
our new models using normal ( i.e. non primordial ) stars show an excellent agreement with the lynx arc measured line ratios , better than the filtered agn models ( see for instance , [ oiii]@xmath65007/hb , nv / hb , civ / h@xmath18 , ciii]/h@xmath18 in table 1 of @xcite ) .
i.e. there is no need for an exotic agn model to reproduce the line ratios .
* the very narrow lines of the lynx arc ( fwhm of non resonant lines @xmath4100 km s@xmath34 , @xcite ) are typical of star forming objects .
agns usually show line widths of few hundred km s@xmath34 .
although there are examples of extended emission line regions in active galaxies with narrower lines , these cases are exceptional . * the strongest discrepancies of the filtered agn model disappear with our new models : these are the severe underprediction of the siiii ] lines and the too high electron temperature ( @xmath920 000 k ) predicted by the filtered agn models .
we therefore reject the possibility that the lynx arc is an active galaxy . _ a hidden starburst _
hidden starbursts have been proposed to explain the spectroscopic properties of the metal poor blue compact galaxy hs 0837 + 4717 ( pustilnik et al .
2004 ) . in this object ,
the extinction of the narrow emission lines from the giant hii region is low , but the large balmer decrement ( @xmath919@xmath104 ) of the broad components suggests that part of the current starburst is highly obscured by dust .
a highly obscured starburst has also been revealed by mid - ir observations of the very low metallicity galaxy sbs 0335 - 052 ( @xmath37=1/41 @xmath38@xcite ) .
roughly 3/4 of the star formation in this object occurs within the obscured cluster .
it is not clear however , that this scenario solves the the lynx arc problem . in hs 0837 + 4717 and
sbs 0335 - 052 the non - obscured regions emit both strong continuum and emission lines .
i.e. , if optical emission lines are detected , the ionizing stars are also detected .
this is not the case for the lynx arc .
_ differential gravitational lensing amplification
_ the different spatial distribution between ionized gas and young stars can be the solution to the lynx continuum problem .
the discrepancy between reddening values derived using the uv continuum and the ratios of the balmer lines found in some star forming galaxies ( e.g. @xcite ) has been explained as due to ionized gas and stars having different spatial distributions and suffering different extinction by dust .
when this discrepancy is found , the uv continuum seems to be less affected by extinction than the emission lines .
the opposite case would apply to the lynx arc .
spatial decoupling between ionized gas and young stars has been observed in some star forming galaxies ( e.g. @xcite ) and proposed by several authors to explain different issues ( e.g. @xcite , @xcite ) the degree of amplification suffered by the source undergoing lensing depends strongly on both the position close to the critical lines and the geometry of the lensed object , so that separate regions can suffer different degrees of amplification . if stars and gas have different spatial distributions in the lynx arc , emission lines and stellar continuum can suffer different magnification due to the intervening cluster .
differential amplification between lines and continuum has been observed in several quasars ( e.g. @xcite ) .
detailed gravitational lensing models would be necessary to verify this possibility . _ a primordial stellar population _
the sed of metal - free stars is characterized by effective temperatures on the main sequence around 10@xmath2 k , hotter than their counterparts of equal mass but finite metallicity ( e.g. @xcite ) .
the uv recombination emission lines , such as heii@xmath61640 , would have extreme equivalent widths more than an order of magnitude larger than the expectation for a normal cluster of hot stars with the same total mass and a salpeter imf .
bromm , kudritzki & loeb ( 2001 ) estimated that a cluster of 10@xmath23 m@xmath11 of popiii stars with a heavy imf would produce @xmath92.3@xmath110@xmath39 erg s@xmath34 in the heii@xmath61640 line and a spectral luminosity per unit frequency of 1.8@xmath110@xmath40 erg s@xmath34 hz@xmath34 . for the observed luminosity of the heii@xmath61640 in the lynx arc ( @xmath94.4@xmath110@xmath39 erg s@xmath34 ) a cluster of @xmath92@xmath110@xmath23 m@xmath11 of primordial stars
would be required .
such a cluster would produce a spectral flux per unit frequency of 3.1@xmath110@xmath33 erg s@xmath34 @xmath35 hz@xmath34 at the distance of the lynx arc .
this is a factor of @xmath92.3 below the continuum measured for the lynx arc .
the discrepancy , therefore , disappears and this result is consistent with the conclusion by @xcite that the continuum is mostly nebular .
the possibility that the lynx arc is a popiii object was proposed and discussed in detail by @xcite .
it is difficult to explain , however , nebular abundances as high as @xmath910% if the stellar population is primordial .
we have characterized the physical properties ( electron temperature , density , chemical abundances ) of the ionized gas and the ionizing stellar population in the lynx arc , a gravitationally amplified hii galaxy at @xmath41=3.36 .
the temperature sensitive ratio [ oiii]@xmath51661,1666/@xmath65007 implies an electron temperature @xmath7=17300@xmath8 k , in good agreement within the errors with photoionization model predictions .
the uv doublets imply the existence of a density gradient in this object , with a highly ionized high density region ( 0.1 - 1.0 @xmath1 10@xmath2 @xmath3 ) and a low density region ( @xmath43200 @xmath3 ) with lower ionization state .
both the photoionization modeling and standard techniques of chemical abundance determination imply that the gas metallicity is @xmath910@xmath103% z@xmath11 .
both methods suggest that nitrogen is overabundant with n / [email protected] - 3@xmath1[n / o]@xmath11 , unless a density gradient produces this apparent effect .
we do not find evidence for si overabundance as fosbury et al .
the reason is the different shape of the ionizing continuum assumed by these authors ( 80 000 k black body , much hotter than the dominant ionizing stars in our models ) .
photoionization models imply that the ionizing stars have very young ages @xmath125 myr .
since the emission lines trace the properties of the present burst only , nothing can be said about the possible presence of an underlying old stellar population .
instantaneous burst models with z@xmath14@xmath155% z@xmath11 and ages @xmath92.8 - 3.4 myr ( depending on z@xmath14 ) , are in excellent agreement with the lynx spectrum , including the strong he@xmath16 emission . at this
age wolf rayet stars make an important contribution to the hard ionizing luminosity and they are responsible for the excitation of the he@xmath16 emission . in such case , we infer the existence of @xmath92.5@xmath110@xmath42 wr stars in the lynx arc .
alternative excitation mechanisms for he@xmath16 , however , can not be discarded .
therefore , the lynx arc is a low metallicity hii galaxy that is undergoing a burst of star formation of @xmath125 myr age .
one possible scenario that explains the emission line spectrum of the lynx arc ,
the strength of the nitrogen lines and the strong he@xmath16 emission is that the object has experienced a merger event that has triggered a powerful starburst phase .
wolf rayet stars have been formed and contribute to a fast chemical enrichment of the interstellar medium .
as fosbury et al .
( 2003 ) , we find a factor of @xmath1510 discrepancy between the mass of the instantaneous burst implied by the luminosity of the h@xmath18 line and the mass implied by the continuum level measured for the lynx arc .
we have discussed several possible solutions to this problem .
the most satisfactory explanation is that gas and stars have different spatial distribution so that the emission lines and the stellar continuum suffer different gravitational amplification by the intervening cluster .
detailed gravitional lensing models are needed to test the vailidity of this scenario .
we thank an anonymous referee for providing very useful comments that helped to improve this paper substantially .
we thank valentina luridiana for useful scientific discussions and andrew humphrey for providing figure 1 .
m. villar - martn and m. cervio are supported by the spanish national program ramn y cajal .
we acknowledge support by the spanish ministry of science and technology ( mcyt ) through grant aya-2001 - 3939-c02 - 01 .
legrand f. , kunth d. , roy j - r .
, mas - hesse j.m . ,
walsh j.r . , 1997 ,
a&a , 326 , l17 leitherer c. , schaerer d. , goldader j.d . , gonzlez delgado r.m . ,
robert c. , foo kune d. , de mello d.f .
, devost d. , heckman t. , 1999 , apjss , 123 , 3 mas - hesse j.m .
, kunth d. , 1998 , in _ ultraviolet astrophysics beyond the iue final archive _ : proceedings of the conference held in sevilla , spain , from 11th to 14th november 1997 . w. wamsteker and r. gonzalez riestra eds .. noordwijk , the netherlands : esa publications division , 1998 .
( esa sp ; 413 ) , p.537 | we have characterized the physical properties ( electron temperature , density , metallicity ) of the ionized gas and the ionizing population ( age , metallicity , presence of wr stars ) in the lynx arc , a hii galaxy at @xmath03.36 .
the uv doublets ( ciii ] , siiii ] and niv ) imply the existence of a density gradient in this object , with a high density region ( 0.1 - 1.0 @xmath1 10@xmath2 @xmath3 ) and a lower density region ( @xmath43200 @xmath3 ) .
the temperature sensitive ratio [ oiii]@xmath51661,1666/@xmath65007 implies an electron temperature @xmath7=17300@xmath8 k , in agreement within the errors with photoionization model predictions .
nebular abundance determination using standard techniques and the results from photoionization models imply a nebular metallicity of o / h@xmath910@xmath103% ( o / h)@xmath11 , in good agreement with fosbury et al ( 2003 ) .
both methods suggest that nitrogen is overabundant relative to other elements , with [ n / o]@xmath92.0 - 3.0 @xmath1 [ n / o]@xmath11 . we do not find evidence for si overabundance , as fosbury et al . ( 2003 ) .
photoionization models imply that the ionizing stellar population in the lynx arc has an age of @xmath125 myr .
if he@xmath13 is ionized by wr stars , then the ionizing stars in the lynx arc have metallicities z@xmath14@xmath155% z@xmath11 and ages @xmath92.8 - 3.4 myr ( depending on z@xmath14 ) , when wr stars appear and are responsible for the he@xmath16 emission .
however , alternative excitation mechanisms for this species are not discarded .
since the emission lines trace the properties of the present burst only , nothing can be said about the possible presence of an underlying old stellar population .
the lynx arc is a low metallicity hii galaxy that is undergoing a burst of star formation of @xmath125 myr age .
one possible scenario that explains the emission line spectrum of the lynx arc , the large strength of the nitrogen lines and the he@xmath16 emission is that the object has experienced a merger event that has triggered a burst of star formation .
wr stars have formed that contribute to a fast enrichment of the ism .
as fosbury et al .
( 2003 ) , we find a factor of @xmath1710 discrepancy between the mass of the instantaneous burst required to power the luminosity of the h@xmath18 line and the mass implied by the continuum level measured for the lynx arc .
we discuss several possible solutions to this problem .
the most likely explanation is that gas and stars have different spatial distributions so that the emission lines and the stellar continuum suffer different gravitational amplification by the intervening cluster .
= = = = = = = = # 1 # 1 # 1 # 1 @mathgroup@group @mathgroup@normal@groupeurmn @mathgroup@bold@groupeurbn @mathgroup@group @mathgroup@normal@groupmsamn @mathgroup@bold@groupmsamn = `` 019 = ' ' 016 = `` 040 = ' ' 336
= " 33e = = = = = = = = # 1 # 1 # 1 # 1 = = = = = = = = [ firstpage ] cosmology : observations galaxies : abundances galaxies : high redshift hii regions stars : formation |
the nuclear structure function @xmath0 is smaller than @xmath1 times the free nucleon structure function @xmath2 for values of @xmath3 in the regime where valence quarks are dominant .
this phenomenon , known as the european muon collaboration ( emc ) effect @xcite , has been known for almost twenty years .
nevertheless , the significance of this observation remains unresolved even though there is a clear interpretation within the parton model : a valence quark in a bound nucleon carries less momentum than a valence quark in a free one .
there are many possible explanations , but no universally accepted one . the underlying mechanism responsible for the transfer of momentum within the constituents of the nucleus
has not yet been specified .
one popular mechanism involves ordinary nuclear binding which , in its simplest inculcation , is represented by evaluating the free nucleon structure function at a value of @xmath3 increased by a factor of the average separation energy divided by the nucleon mass @xmath4 .
the validity of this binding effect has been questioned ; see the reviews @xcite .
the bjorken variable @xmath3 is a fraction of the plus component of momentum , and the desire to obtain a more precise evaluation and understanding of the binding effect lead us to attempt to obtain a nuclear wave function in which the momentum of the nucleons is expressed in terms of this same plus component .
therefore we applied light front dynamics to determining nuclear wave functions @xcite . in this formalism
one defines @xmath5 and quantizes on equal @xmath6 surfaces which have a constant light front time , @xmath7 .
the conjugate operator @xmath8 acts as an evolution operator for @xmath7 .
the plus momentum is canonically conjugate to the spatial @xmath9 variable .
this light front formalism has a variety of advantages @xcite and also entails complications @xcite .
our most recent result @xcite is that the use of the relativistic mean field approximation , and the assumption that the structure of the nucleon is not modified by effects of the medium , to describe infinite nuclear matter leads to no appreciable binding effect .
the failure was encapsulated in terms of the hugenholtz - van hove theorem @xcite which states that the average nuclear binding energy per nucleon is equal to the binding energy of a nucleon at the top of the fermi sea .
the light front version of this theorem is obtained from the requirement that , in the nuclear rest frame , the expectation values of the total plus and minus momentum are equal .
the original version of the theorem was obtained in a non - relativistic theory in which nucleons are the only degrees of freedom . here , the mesons are important and the theory is relativistic , but the theorem still holds . this theorem can be shown to restrict @xcite the plus momentum carried by nucleons to be the mass of the nucleus , which in turn implies that the probability for a nucleon to have a plus momentum @xmath10 is narrowly peaked about @xmath11 .
thus the only binding effect arises from the average binding energy , which is much smaller than the average separation energy .
therefore dynamics beyond the relativistic mean field approximation must be invoked to explain the emc effect .
this conclusion was limited to the case of infinite nuclear matter , and the computed nuclear structure function could only be compared with data on finite nuclear targets extrapolated to the limit @xmath12 . the goal of the present work is to extend the results to finite nuclei ; the main complication arises from the spatial dependence of the nucleon and meson fields .
we briefly outline our procedure . in sections
[ sec : green ] and [ sec : subtle ] we present the covariant deep inelastic scattering formalism of ref .
@xcite and derive its representation in terms of nucleon single particle wave functions .
the plus momentum distribution follows from this representation in section [ sec : derivation ] where we also derive new version of the hugenholtz - van hove theorem .
then we present the results of analytic and numerical calculations in section [ sec : numbers ] , the latter giving an @xmath1 dependence of the ratio function contrary to experimental results .
this again gives the result that the use of the relativistic mean field approximation , combined with the assumption that the nuclear medium does not modify the structure of the nucleon , can not describe the emc effect .
the reasons for the subtle differences between the results for finite nuclei and nuclear matter are detailed in section [ sec : scalar ] .
finally , we summarize and discuss possible implications .
we begin with the covariant plus momentum distribution function @xmath13 \label{eq : fn}\ ] ] where we identify @xmath14 where @xmath15 is the connected part of the nucleon green s function : @xmath16 this result is directly determined from the feynman diagram in fig .
[ fig : dis ] is struck by a virtual photon of momentum @xmath17 .
we label nucleon momentum @xmath18 , and quark momentum @xmath19 . ]
following ref .
so far this is independent of the particular relativistic mean field model , but for concreteness we use a quantum hadrodynamics ( qhd ) lagrangian @xcite , specifically qhd - i as in ref .
@xcite , where the nucleon fields , @xmath20 , that appear in eq .
( [ eq : green ] ) are those appearing in the lagrangian .
light front quantization requires that the plus component of all vector potential fields vanishes , and this is obtained by using the soper - yan transformation @xcite @xmath21 to define the nucleon field operator @xmath22 for various models @xcite .
this transformation allows the use of the eigenmode expansion for the @xmath22 fields which have been obtained previously in ref .
@xcite @xmath23\nonumber \\
& = & \sum_{\alpha } \left [ a_{\alpha } u_{\alpha } ( \bm{x } ) e^{-i p^{-}_{\alpha } x^{+}/2 } + b_{\alpha}^{\dag } v_{\alpha } ( \bm{x } ) e^{i p^{-}_{\alpha } x^{+}/2 } \right ] , \label{eq : mode - exp}\end{aligned}\ ] ] where @xmath24 and @xmath25 are ( anti-)nucleon annihilation operators and we define @xmath26 with @xmath27 and @xmath28 which allows us to treat the minus and perpendicular cordinates on equal footing .
the @xmath29 and @xmath30 are cordinate space 4-component spinor solutions to the light front dirac equation with eigenvalues @xmath31 . to simplify the analysis
we will temporarily ignore electromagnetic effects , but we will include them in the final numerical results .
the light front mode equations in qhd - i are obtained by minimizing the @xmath8 operator ( light front hamiltonian ) with the constraint @xcite that @xmath32 .
the result is @xmath33 p^{-}_{\alpha}|\psi_{\alpha}^{+}\rangle & = & \left[-i\partial_{z}+2g_{v}\bar{v}^{-}\right]|\psi_{\alpha}^{+}\rangle \nonumber\\ * & & + \left[\bm{\alpha}_{\bot}\cdot(\bm{p}_{\bot}-g_{v}\bar{\bm{v}}_{\bot})+\beta(m+g_{s}\phi)\right ] @xmath34 using standard manipulations @xcite and defining @xmath35 as the energy of the highest occupied state , we find the green s function to be @xmath36 } \int \frac{dk^{-}}{2\pi } e ^{-ik^{-}(x^{+}-y^{+})/2 } \left [ \frac{1}{k^{-}-p^{-}_{\alpha } + i\varepsilon } + 2\pi i \delta(k^{-}-p^{-}_{\alpha } ) \theta(\varepsilon_{f}-\varepsilon_{\alpha } ) \right]\nonumber\\ & \equiv & g^{d}(x , y ) + g^{c}(x , y),\end{aligned}\ ] ] where the superscripts @xmath37 and @xmath38 represent the disconnected and connected parts of the nucleon green s function , respectively .
the connected part is relevant to deep inelastic scattering and is given by @xmath39 } e^{-ip^{-}_{\alpha}(x^{+}-y^{+})/2 } , \label{eq : cgreen}\ ] ] where the sum is over occupied levels @xmath40 in the fermi sea @xmath41 .
we now substitute eq .
( [ eq : cgreen ] ) into eq .
( [ eq : chi ] ) , first defining @xmath42 where @xmath43 , @xmath44 , @xmath45 and @xmath46 we find @xmath47 the motivation for the ` double - prime ' notation is the subject of the next section .
it would be useful to express @xmath48 in terms of solutions of the ordinary dirac equation , because one may use a standard computer program @xcite . to this end
it is convenient rewrite eq .
( [ eq : uprime ] ) @xmath49 of eq . ( [ eq : tranpsi ] ) and @xmath50 of eq .
( [ eq : uprime ] ) is simply the normalization factor @xmath51 .
these ` double - primed ' fields satisfy another version of the mode equations eq .
( [ eq : dirac1 ] ) and eq .
( [ eq : dirac2 ] ) following from an application of the soper - yan transformation eq .
( [ eq : soperyan ] ) , and are given by @xmath52|\psi_{\alpha}''^{-}\rangle & = & \left[\bm{\alpha}_{\bot}\cdot(\bm{p}_{\bot } -g_{v}\bm{v}_{\bot } ) + \beta(m+g_{s}\phi)\right ] \left[i\partial_{z}+g_{v}v^{+}-2g_{v}\bar{v}^{-}+p^{-}_{\alpha}\right]|\psi_{\alpha}''^{+}\rangle & = & \left[\bm{\alpha}_{\bot}\cdot(\bm{p}_{\bot } -g_{v}\bm{v}_{\bot } ) + \beta(m+g_{s}\phi)\right ] one multiplies eq .
( [ eq : dirac1-primed ] ) by @xmath53 and eq .
( [ eq : dirac2-primed ] ) by @xmath54 and adds the two equations , using @xmath55 , one obtains @xmath56 \psi_{\alpha}''(\bm{x } ) = \left [ \bm{\gamma}_{\bot}\cdot\bm{p}_{\bot } + m+g_{s}\phi \right]\psi_{\alpha}''(\bm{x } ) \label{eq : almost - dirac}\ ] ] eq .
( [ eq : almost - dirac ] ) is almost the same as the dirac equation of the equal time formulation ( for the @xmath20 fields ) , with the exception of the @xmath57 term multiplying @xmath58 .
removing the offending term gives the relationship between the @xmath20 and @xmath59 fields @xmath60 eq . ( [ eq : prime - noprime ] ) is the approximate relationship between the @xmath22 and @xmath20 fields determined in ref .
the approximation lies in the fact that the spectrum condition is not maintained exactly , and the resulting fourier transform of the wavefunction will have unphysical support for @xmath61 .
this support is largely irrelevant as it manifests far out on an exponential tail since @xmath57 contains the nucleon mass .
the relationship between the @xmath22 and @xmath59 is exactly our definition eq .
( [ eq : tranpsi ] ) . the use of eq .
( [ eq : prime - doubleprime ] ) in eq .
( [ eq : almost - dirac ] ) leads immediately to the result that the fields @xmath62 satisfy the ordinary dirac equation @xmath63\psi_{\alpha}'(\bm{x})\ ] ] we now are ready to derive a representation of eq .
( [ eq : fn ] ) in terms of these nucleon wave functions .
@xcite , it was determined that a plus momentum distribution in qhd - i is given by @xmath64 this distribution peaks at @xmath65 for @xmath66o , ( with smaller values for heavier nuclei ) but is not the distribution obtained from the covariant formalism of section [ sec : green ] .
the connection between this @xmath67 and the covariant @xmath68 was made in ref .
@xcite ; it was determined that , in the limit of infinite nuclear matter , the relationship between @xmath67 and @xmath68 is simply a shift in the argument by the vector meson potential : @xmath69 this shift arises from the use of the soper - yan transformation eq .
( [ eq : soperyan ] ) where the @xmath20 fields are those appearing in the lagrangian and are used to determine @xmath68 , whereas the @xmath22 fields are used to determine @xmath67 . in finite nuclei , this relationship is somewhat more complicated since the vector meson potential is no longer a constant over all space .
we start with eq .
( [ eq : new - chi ] ) , and see that @xmath70 \delta(k^{-}-p^{-}_{\alpha})\nonumber\\ * & = & 16\pi^{2 } p ^{+ } a \sum_{\alpha\in f } \left|\langle \bm{k}|\psi''^{+}_{\alpha}\rangle \right|^{2 } \delta(k^{-}-p^{-}_{\alpha}).\nonumber\end{aligned}\ ] ] substituting into eq .
( [ eq : fn ] ) we obtain latexmath:[\ ] ] the gaussian parameterization uses the peak location and width , @xmath108 and @xmath122 respectively , from the relativistic hartree calculations in fig .
[ fig : fn ] , and is normalized to unity .
this allows us to obtain a plus momentum distribution for any @xmath1 with minimal effort .
we show the combined effect of scalar mesons and binding energy per nucleon on the ratio function along with the effect of scalar mesons alone using a constant binding energy per nucleon of @xmath120 independent of @xmath1 .
it can be seen that a changing @xmath115 with @xmath1 has the most effect for nuclei much larger than iron , but does not change the general trend that the minimum of the ratio function becomes less pronounced as @xmath1 increases due to the vanishing scalar meson contribution and the peak of the plus momentum distribution approaching unity .
this dependence of the binding effect on @xmath1 is quite different , both in magnitude and shape , than the trend in experimental data summarized in ref .
@xcite which satisfies @xmath123 , so that the minimum becomes more pronounced as @xmath1 increases .
this fully demonstrates the inadequacy of conventional nucleon - meson dynamics to explain the emc effect .
the average value of @xmath73 , given by eq .
( [ eq : hvh ] ) , yields the nucleon contribution to the plus momentum , and is less than one which can be seen in fig . [ fig : fn ] .
we now address the remaining plus momentum in finite nuclei .
previous results @xcite show that a small fraction ( @xmath124 ) of the plus momentum is carried by the scalar mesons which vanishes as the nuclear matter limit is approached .
this is due to the fact that scalar mesons couple to gradients in the scalar density ( arising mainly from the surface of finite nuclei ) which vanish as @xmath114 .
the question is : why are scalar mesons allowed to carry plus momentum and not vector mesons ?
the simplest answer lies in the dirac structure of eq .
( [ eq : fn ] ) ; the @xmath53 in the trace picks out terms in the full interacting green s function with an odd number of gamma matrices which includes all lorentz vector interactions and excludes lorentz scalar interactions .
the dirac structure of @xmath68 is directly related to the dirac structure of the energy momentum tensor , so the answer also lies there and illuminates a problem with conventional nucleon - meson dynamics .
the component of the energy momentum tensor relevant to the plus momentum , from a chiral lagrangian containing isoscalar vector mesons , scalar mesons and pions , is given by @xcite @xmath125 since each of the terms in eq .
( [ eq : t++ ] ) involves one of the fields , it is natural to associate each term with a particular contribution to the plus momentum .
this decomposition , though , is not well defined ; field equations relate various components .
we see the first three terms of eq .
( [ eq : t++ ] ) appear in @xmath126 , which defines the nucleon contribution to the total nuclear plus momentum , in the derivation of the hugenholtz - van hove theorem eq .
( [ eq : hvh ] ) ; we are not allowed to have the vector mesons contribute a well defined fraction of plus momentum .
this means that one could trade certain mesonic degrees of freedom for nucleons by replacing mesonic vertices with nucleon point couplings , for example , in line with the general concept of effective field theory . in our case
the first three terms are related by the vector meson field equation , but the fourth is left out since the scalar mesons couple to the scalar density @xmath127 which is not present in eq .
( [ eq : t++ ] ) . therefore the scalar mesons ( and pions )
contribute a well defined fraction of plus momentum .
these explicit meson contributions create an emc binding effect , but the pionic contributions are also limited by nuclear drell - yan experiments @xcite to carrying about 2% of the plus momentum which is insufficient to account for the entire emc effect which corresponds to about 5% of the plus momentum for iron .
the minimum in the emc effect is known to have a monotonically decreasing behavior with @xmath1 , which has been studied in refs .
@xcite among others .
our present theory is defined by the use of the mean - field approximation , along with the assumption that nuclear effects do not modify the structure of the nucleon .
this theory leads to results in severe disagreement with experiment . not only
do we find that the depth of the minimum is monotonically decreasing with @xmath1 , but it has a smaller magnitude than experiment .
these results , which fail to capture any of the important features of the experiments , represent a failure of relativistic mean field theory .
furthermore , the plus momentum distributions we compute give @xmath128 which indicates that nearly all of the plus momentum is carried by the nucleons . in order to reproduce the data ,
the nucleon plus momentum must be decreased by some mechanism that becomes more important at larger @xmath1 .
nucleon - nucleon correlations can not take plus momentum from nucleons , and explicit mesonic components in the nuclear fock state wavefunction carrying plus momentum are limited @xcite by drell - yan experiments @xcite .
thus it appears that the emc effect may be due to something outside of conventional nucleon - meson dynamics .
for example , true modifications to nucleon structure caused by nuclear interactions could be important , in which case one would need to use models such as the mini - delocalization model @xcite , quark - meson coupling ( qmc ) model @xcite or the chiral quark soliton model reviewed in @xcite to include those effects .
j. j. aubert _ et al .
_ [ european muon collaboration ] , phys .
b * 123 * , 275 ( 1983 ) .
m. arneodo , phys .
rept . * 240 * , 301 ( 1994 ) .
d. f. geesaman , k. saito and a. w. thomas , ann .
nucl . part .
* 45 * , 337 ( 1995 ) .
g. piller and w. weise , phys .
* 330 * , 1 ( 2000 ) [ hep - ph/9908230 ] .
l. l. frankfurt and m. i. strikman , phys .
rept . * 160 * , 235 ( 1988 ) .
g. a. miller , prog .
phys . * 45 * , 83 ( 2000 ) .
s. j. brodsky , h. c. pauli , and s. s. pinsky , phys .
rept . * 301 * , 299 ( 1998 ) .
r. j. perry , in `` hadron physics 94 : topics on the structure and interactions of hadronic systems '' , v. e. herscovitz _ et al_. ( eds . ) , world scientific , singapore ( 1994 ) [ hep - th/9407056 ] .
a. harindranath , in `` light - front quantization and non - perturbative qcd '' , j.p . vary and f. wlz ( eds . ) , int . inst . of theoretical and applied physics ( 1997 ) [ hep - ph/9612244 ] .
j. r. cooke and g. a. miller , [ nucl - th/0112037 ] .
g. a. miller and j. r. smith , phys .
c * 65 * , 015211 ( 2002 ) [ nucl - th/0107026 ] .
g. a. miller , phys .
c * 56 * , 2789 ( 1997 ) [ nucl - th/9706028 ] .
g. a. miller and r. machleidt , phys .
c * 60 * , 035202 ( 1999 ) [ nucl - th/9903080 ] .
h. jung and g. a. miller , phys .
b * 200 * , 351 ( 1988 ) .
g. blunden , m. burkardt and g. a. miller , phys .
c * 60 * , 055211 ( 1999 ) [ nucl - th/9906012 ] .
b. d. serot and j. d. walecka , int .
j. mod .
e * 6 * , 515 ( 1997 ) [ nucl - th/9701058 ] .
b. d. serot and j. d. walecka , adv .
* 16 * , 1 ( 1986 ) .
d. e. soper , phys .
d * 4 * , 1620 ( 1971 ) .
yan , phys .
d * 7 * , 1760 ( 1973 ) .
m. c. birse , phys .
b * 299 * , 186 ( 1993 ) . c. j. horowitz and b. d. serot , nucl .
a * 368 * , 503 ( 1981 ) . `
timora ` by c. j. horowitz , in `` computational nuclear physics .
vol . 1 : nuclear structure '' k. langanke , j. a. maruhn , and s. e. koonin ( eds . ) , springer - verlag , new york ( 1991 ) . j. gomez _ et al .
_ [ slac - e139 ] , phys .
d * 49 * , 4348 ( 1994 ) .
j. g. de groot _
_ , phys .
b * 82 * , 456 ( 1979 ) .
i. sick and d. day , phys .
b * 274 * , 16 ( 1992 ) .
r. p. bickerstaff , m. c. birse and g. a. miller , phys .
* 53 * , 2532 ( 1984 ) .
m. ericson and a. w. thomas , phys .
b * 148 * , 191 ( 1984 ) .
e. l. berger , nucl .
b * 267 * , 231 ( 1986 ) .
d. m. alde _ et al .
_ , phys .
* 64 * , 2479 ( 1990 ) .
k. saito and a. w. thomas , phys .
b * 327 * , 9 ( 1994 ) [ nucl - th/9403015 ] .
o. benhar , v. r. pandharipande and i. sick , phys .
b * 469 * , 19 ( 1999 ) .
f. gross and s. liuti , phys .
c * 45 * , 1374 ( 1992 ) .
d. diakonov and v. y. petrov , [ hep - ph/0009006 ] . | a light front formalism for deep inelastic lepton scattering from finite nuclei is developed . in particular , the nucleon plus momentum distribution and a finite system analog of the hugenholtz - van hove theorem are presented . using a relativistic mean field model , numerical results for the plus momentum distribution and ratio of bound to free nucleon structure functions for oxygen ,
calcium and lead are given .
we show that we can incorporate light front physics with excellent accuracy while using easily computed equal time wavefunctions .
assuming nucleon structure is not modified in - medium we find that the calculations are not consistent with the binding effect apparent in the data not only in the magnitude of the effect , but in the dependence on the number of nucleons .
5.5innt@uw-02 - 001 |
the latest astronomical observations have given a substantial boost to the development of modern cosmology @xcite .
in particular , the accelerating expansion of the universe has drawn much attention .
the early time acceleration is widely known as _ inflation _ , while the late time acceleration is usually referred to as the epoch of _ dark energy _
presently , the @xmath0cdm model , in which the cosmological constant @xmath0 plays the role of dark energy , is accepted as a standard cosmological model .
there is an extensive literature on other forms of dark energy , too @xcite .
all in all , the number of dark energy models that can be found in literature is enormous .
the same holds for the inflationary models that have been constructed over the years . in this paper
, i shall describe the procedure which associates an action functional with each chosen geometry of the universe .
precisely , any desirable geometry of the universe is made a solution of a particular geometric sigma model .
geometric sigma models are theories that possess two distinctive properties .
first , their complete matter content can be gauged away .
second , any predefined geometry can be made a solution of a particular model .
these models have first been proposed in @xcite in the context of fermionic excitations of flat geometry .
here , i shall use them for modeling the dynamics of the universe .
this way , the variety of existing cosmological models in literature is replaced with a single concept . the results obtained in this paper are summarized as follows .
first , a class of purely geometric theories of the universe has been defined by associating a geometric sigma model to every spatially flat , homogeneous and isotropic geometry .
this way , each geometry of this kind becomes a vacuum solution of a particular geometric sigma model .
ultimately , one is provided with the action functional from which the observed cosmology can be derived .
the inflation and the late time acceleration have purely geometric origin .
the second result concerns the vacuum stability of geometric sigma models .
it has been proven true for almost all background geometries .
this is an important achievement because the stability is not guaranteed by the very construction of geometric sigma models . instead
, the dynamics of small perturbations has been examined by direct calculation .
the layout of the paper is as follows . in sec .
[ sec2 ] , the construction of geometric sigma models , as defined in @xcite , is recapitulated and subsequently applied to spatially flat , homogeneous and isotropic geometries . as a result ,
a class of action functionals of the universe is obtained .
each of these action functionals possesses a topologically nontrivial vacuum solution that describes a particular universe . in sec .
[ sec3 ] , the dynamics of small perturbations of the topologically nontrivial vacuums is examined . in sec .
[ sec4 ] , the vacuum solutions are proven stable for all spatially flat , homogeneous and isotropic background geometries . in sec .
[ sec5 ] , the examples of inflationary and standard model universes are used to demonstrate how geometric sigma models are constructed in practice . sec .
[ sec6 ] is devoted to concluding remarks .
my conventions are as follows .
the indexes @xmath1 , @xmath2 , ... and @xmath3 , @xmath4 , ... from the middle of the alphabet take values @xmath5 .
the indexes @xmath6 , @xmath7 , ... and @xmath8 , @xmath9 , ... from the beginning of the alphabet take values @xmath10 .
the spacetime coordinates are denoted by @xmath11 , the ordinary differentiation uses comma ( @xmath12 ) , and the covariant differentiation uses semicolon ( @xmath13 ) .
the repeated indexes denote summation : @xmath14 .
the signature of the @xmath15-metric @xmath16 is @xmath17 , and the curvature tensor is defined as @xmath18 .
the construction of a geometric sigma model begins with specifying a spacetime metric .
i shall denote it with @xmath19
. the metric @xmath20 is freely chosen , and the coordinates @xmath11 are fully fixed . as a consequence ,
the functions @xmath19 are completely determined . in the next step ,
the corresponding ricci tensor @xmath21 is calculated , and the following einstein like equation is postulated : @xmath22 if @xmath23 , the metric @xmath20 is a solution of the free einstein s equation .
if , on the other hand , @xmath24 , the non - zero right hand side of the equation ( [ 1 ] ) _ defines _ matter content of the theory .
the equation ( [ 1 ] ) obviously lacks general covariance . to covariantize it
, i introduce a new set of coordinates @xmath25 . in terms of these new coordinates ,
the equation ( [ 1 ] ) takes the form @xmath26 where the functions @xmath27 are defined through @xmath28 the equation ( [ 2 ] ) is generally covariant once the new coordinates @xmath29 are seen as scalar functions of the old coordinates @xmath11 .
moreover , the equation ( [ 2 ] ) reduces to ( [ 1 ] ) in the coordinate frame defined by @xmath30 .
the equation ( [ 2 ] ) has the form of the einstein s equation in which four scalar fields @xmath31 of some nonlinear sigma model are coupled to gravity .
the `` matter field equations '' are obtained by utilizing the bianchi identities .
if the condition @xmath32 is fulfilled , one obtains @xmath33 it is straightforward to verify that the equations ( [ 2 ] ) and ( [ 4 ] ) can be derived from the action functional @xmath34 \,.\ ] ] it describes a nonlinear sigma model coupled to gravity , and possesses the topologically nontrivial ( let me call it _ vacuum _ ) solution @xmath35 the physics of small perturbations of the vacuum ( [ 6 ] ) does not violate the condition @xmath32 , which enables one to interpret the scalars @xmath29 as spacetime coordinates . if this is the case , one is allowed to fix the gauge @xmath36 , which brings us back to the geometric equation ( [ 1 ] ) .
one should have in mind , however , that the action ( [ 5 ] ) has a non - geometric sector , too .
it is characterized by @xmath37 , and includes topologically trivial solutions such as @xmath38 . in what follows
, i shall restrict my considerations to purely geometric dynamics of small perturbations of the vacuum ( [ 6 ] ) . before i continue , let me note that the equation ( [ 1 ] ) is not the unique geometric equation that allows the solution @xmath39 .
a simple generalization of this equation can be obtained by adding terms proportional to @xmath40 .
the simplest choice is the equation @xmath41 it defines a class of geometric theories parametrized by metrics @xmath20 , and potentials @xmath42 .
the covariantization of the non - covariant equation ( [ 7 ] ) ultimately leads to the action functional @xmath43 \,,\ ] ] where the target metric @xmath44 is defined by @xmath45 the class of theories defined by ( [ 8 ] ) possesses the vacuum solution ( [ 6 ] ) for any choice of the potential @xmath46 .
the physics of small perturbations of this vacuum allows the gauge condition @xmath47 , which brigs us back to the geometric equation ( [ 7 ] ) . in what follows
, i shall associate a geometric sigma model to a given geometry of the universe .
let me choose a spatially flat , homogeneous and isotropic metric @xmath48 the corresponding ricci tensor is calculated straightforwardly .
one obtains @xmath49 where `` dot '' denotes time derivative .
now , i am ready to construct the target metric @xmath44 . by inspecting the definition ( [ 9 ] ) ,
one notices that the target metric @xmath50 can significantly be simplified by a proper choice of the potential @xmath42 .
indeed , if the potential is chosen to have the form @xmath51 the component @xmath52 remains the only nonzero component of the target metric .
precisely , one obtains @xmath53 where the identification @xmath54 is introduced for convenience . as a consequence , @xmath55 is the only scalar field that enters the action functional ( [ 8 ] ) .
let me simplify the notation by using the identification @xmath56 .
the action ( [ 8 ] ) then reduces to @xmath57 .\ ] ] it governs the dynamics of gravity coupled to a scalar field , and possesses the vacuum solution @xmath58 irrespectively of the choice of the scale factor @xmath59 .
the precise form of the target metric @xmath60 and the potential @xmath46 is determined once the function @xmath59 is specified .
the standard physical requirements that ensure the absence of ghosts and tachions restrain the target metric @xmath60 to be positively definite , and the potential @xmath46 to be bounded from below .
these restrictions , however , refer to the topologically trivial vacuum .
precisely , the positive definiteness of @xmath60 , and the fact that @xmath61 is a minimum of the potential @xmath46 ensure stability of the vacuum @xmath61 , @xmath62 . in this paper , however , the vacuum of interest is the topologically nontrivial vacuum ( [ 15 ] ) .
its stability is not guaranteed by the above physical requirements , and i am led to check it by direct calculation .
in this section , i shall examine the dynamics of small perturbations of the vacuum ( [ 15 ] ) , as governed by the action functional ( [ 14 ] ) .
the infinitesimal change of coordinates @xmath63 leaves this action invariant , and allows the gauge fixing @xmath64 . in this gauge ,
the matter field equation is identically satisfied , and we are left with the gravitational field equation ( [ 7 ] ) .
the residual diffeomorphisms are defined by the constraint @xmath65 .
the only variable in the gauge fixed theory is the metric perturbation @xmath66 , defined by @xmath67 with respect to the residual diffeomorphisms , it transforms as @xmath68 where @xmath69 is the form variation , and @xmath70 .
it is seen that @xmath71 can also be gauged away . the gauge condition @xmath72 restrains the gauge parameters to be functions of spatial coordinates , only .
precisely , the residual gauge parameters are defined by @xmath73 in what follows , i shall demonstrate how the residual gauge symmetry can further be fixed .
let me first linearize the field equations ( [ 7 ] ) .
after cumbersome , but straightforward , calculation one obtains [ 20 ] @xmath74 @xmath75 - 2\frac{\dot a}{a}\,h_{00,\alpha } = 0 \,,\ ] ] @xmath76 it is immediately seen that the equation ( [ 20b ] ) implies @xmath77 = 0 \,,\ ] ] which tells us that the expression in square brackets does not depend on time . as a consequence , this expression can be gauged away .
indeed , its transformation law reads @xmath78
= \xi_{\alpha,\gamma\gamma\beta}- \xi_{\beta,\gamma\gamma\alpha } \,.\ ] ] both , the expression in square brackets and the residual gauge parameters @xmath79 are functions of spatial coordinates alone .
this allows the gauge fixing @xmath80 in what follows , i shall simplify the analysis by the assumption that the metric perturbations are _ spatially localized_. this means that the perturbation @xmath66 is assumed to decrease sufficiently fast in spatial infinity . with this assumption ,
many expressions are simplified .
for example , the equation @xmath81 has the general solution @xmath82 , but the adopted boundary conditions imply @xmath83 .
similarly , the equation @xmath84 has the unique solution @xmath83 . with this in mind ,
one easily determines the residual symmetry after the gauge condition ( [ 23 ] ) has been imposed .
it is defined by @xmath85 where the new parameter @xmath86 is an arbitrary function of spatial coordinates .
let me now extract the divergence free parts of the variable @xmath87 . to this end
, i use the decomposition @xmath88 where @xmath89 and @xmath90 are , by definition , divergence free ( @xmath91 ) . in what follows ,
all the expressions will be rewritten in terms of the new variables @xmath92 , @xmath90 and @xmath93 .
let me start with the gauge condition ( [ 23 ] ) . with the help of the adopted boundary conditions , it is straightforward to verify that ( [ 23 ] ) becomes @xmath94 now , i am ready to rewrite the equations ( [ 20 ] ) in terms of the remaining variables @xmath95 , @xmath89 and @xmath93 .
the first two equations take the form [ 27 ] @xmath96_{,0 } + \frac{\dot a}{a } \left [ \frac{1}{a^2 } \tilde h_{\alpha\alpha } \right]_{,\beta\beta } = \,2\dot a^2 \left [ \frac{1}{a^2 } \tilde h \right]_{,0\beta\beta } \,,\ ] ] @xmath97_{,0 } + 2\frac{\dot a}{a}\ , h_{00 } = 0 \,.\ ] ] the third equation ( [ 20c ] ) is too cumbersome to be displayed as a whole .
instead , i shall take its trace and its divergence to obtain two useful scalar equations .
they read [ 28 ] @xmath98_{,0 } + \frac{\dot a^2}{a^2 } \left [ \frac{1}{a^2 } \tilde h_{\alpha\alpha}\right]_{,\beta\beta } = 0 \,,\ ] ] @xmath99_{,00 } + \left(a\dddot a + 3\dot a \ddot a - 3\dot a^2\frac{\dot a}{a } - \ddot a^2 \frac{a}{\dot a}\right ) \left [ \frac{1}{a^2 } \tilde h_{\alpha\alpha } \right]_{,0 } = 0 \,.\ ] ] the equations ( [ 27a ] ) and ( [ 28a ] ) imply @xmath100 , which turns into @xmath101 once the adopted boundary conditions are taken into account .
thus , the variable @xmath102 is independent of time . as a consequence
, the remaining gauge parameter @xmath103 may be used to gauge it away .
indeed , the transformation law of @xmath102 has the form @xmath104 which allows the gauge fixing @xmath105 with this gauge fixing , we are left with no free gauge parameters . nevertheless , the field equations can further be simplified .
take , for example , the equation ( [ 28b ] ) .
if @xmath106 , it is shown to be identically satisfied . if , on the other hand , @xmath107 , it is readily solved .
indeed , the general solution of ( [ 28b ] ) is easily found to be @xmath108 where @xmath109 and @xmath110 are arbitrary functions of spatial coordinates , and @xmath111 is defined by @xmath112 the equation ( [ 28a ] ) then implies @xmath113 for a generic choice of @xmath59 , the functions of time that multiply @xmath109 and @xmath114 are linearly independent .
this leads to @xmath115 , once the assumption of localized perturbations is taken into account .
the equations ( [ 31 ] ) and ( [ 27b ] ) then tell us that @xmath116 the only remaining variable is the traceless and divergenceless quantity @xmath89 . in terms of @xmath92 ,
the cumbersome equation ( [ 20c ] ) takes the simple form @xmath117_{,00 } + 3\frac{\dot a}{a}\left[\frac{1}{a^2 } \tilde h_{\alpha\beta}\right]_{,0 } - \frac{1}{a^2}\left[\frac{1}{a^2 } \tilde h_{\alpha\beta}\right]_{,\gamma\gamma } = 0 \,.\ ] ] the variable @xmath89 is subject to the constraints @xmath118 . owing to this ,
our final equation ( [ 33 ] ) describes a theory with two physical degrees of freedom .
in this section , i shall examine the stability of the vacuum solution @xmath119 of the equation ( [ 33 ] ) .
let me start with the introduction of the collective variable @xmath120 in terms of @xmath121 , the equation ( [ 33 ] ) takes the compact form @xmath122 the function @xmath123 is searched for in the form @xmath124 whereupon the equation ( [ 35 ] ) becomes @xmath125 the stability of the vacuum solution @xmath126 is examined by the canonical analysis of the equation ( [ 37 ] ) . in the first step ,
i notice that the equation ( [ 37 ] ) is obtained from the lagrangian @xmath127 indeed , it is easily verified that its variation leads to the equation ( [ 37 ] ) .
the corresponding hamiltonian is straightforwardly calculated to be @xmath128 it is seen that the hamiltonian is positive for all the allowed values of the parameter @xmath59 , and all the values of the wave vector @xmath129 .
its minimum is located at @xmath130 . for hamiltonians with no explicit time dependence
, this would imply the stability of the vacuum @xmath130 .
indeed , owing to @xmath131 , the physical phase space trajectories coincide with the orbits @xmath132 .
these orbits , on the other hand , are closed curves around the vacuum @xmath130 . as a consequence , the phase space trajectory which is initially close to the vacuum continues to be in the vicinity of the vacuum at all times .
unfortunately , the hamiltonian ( [ 39 ] ) depends on time through the free parameter @xmath59 . as a consequence , @xmath133 and the hamiltonian is not conserved .
still , the stability of the vacuum @xmath130 is not compromised . to see this , note that the orbits @xmath134 remain to be closed curves around @xmath130 , only this time they evolve in time as shown in fig .
[ f1 ] .
.[f1 ] , height=94 ] it is seen from ( [ 40 ] ) that the change of the hamiltonian along the phase space trajectory @xmath135 is the same as for the still point @xmath136 .
thus , the general solution of ( [ 40 ] ) is the phase space trajectory which , at any time , touches the respective orbit @xmath132 .
as these orbits are closed curves around @xmath130 , the phase space trajectories initially close to @xmath130 remain to be close to @xmath130 at all times . in other words ,
the vacuum solution @xmath130 is stable against small perturbations .
this holds true for any @xmath59 and @xmath129 .
in this section , i shall analyze two specific choices of the universe dynamics . in the first , the inflationary epoch of the universe is considered . in the second ,
i propose a model close to the standard model of the universe .
neither of these models is expected to be genuinely realistic .
they are solely used for the demonstration of how an action functional is associated with each chosen universe dynamics .
let me consider a homogeneous , isotropic and spatially flat geometry ( [ 10 ] ) with the scale factor of the simple form @xmath137 its graph is displayed in fig .
it describes an ever existing universe , whose inflationary epoch begins at infinite past and lasts , height=151 ] infinitely many @xmath138-folds .
a graceful exit from inflation happens at @xmath139 .
the constant @xmath140 is a free parameter of the model . in a realistic theory ,
it is determined by astronomical measurements . the scale factor ( [ 41 ] ) is a solution of the sigma model ( [ 14 ] ) in which the potential @xmath46 and the target metric @xmath60 are calculated from ( [ 12 ] ) and ( [ 13 ] ) .
this procedure straightforwardly leads to @xmath141 with @xmath60 and @xmath46 defined by ( [ 42 ] ) , the action functional ( [ 14 ] ) has the solution ( [ 15 ] ) , in which @xmath142 is defined by ( [ 10 ] ) and ( [ 41 ] ) .
the model can further be simplified by a suitable redefinition of the scalar field .
specifically , the redefinition @xmath143 of the form @xmath144 brings the action ( [ 14 ] ) to the form @xmath145 .\ ] ] the new potential @xmath146 reads @xmath147 in terms of @xmath148 , the topologically nontrivial vacuum solution @xmath149 becomes @xmath150 while the metric @xmath151 remains the same .
the graphs of the potential @xmath152 , and the solution ( [ 46 ] ) are displayed in figs .
[ f3 ] and [ f4 ] .
.[f3 ] , height=151 ] it is seen that the potential @xmath152 is a periodic function of @xmath148 , with the period @xmath153 .
thus , the scalar field @xmath148 lives on a circle . the soliton solution ( [ 46 ] ) is one - to - one mapping @xmath154 . as for the trivial solutions , the theory accommodates two of them .
the first is given by @xmath155 , and anti - de sitter metric .
the second has @xmath156 , and the metric is de sitter .
, height=170 ] only the first solution is stable , because @xmath155 is the minimum of the potential @xmath152 , and anti - de sitter metric is a stable solution of the corresponding geometric equation @xcite .
the unstable solution @xmath157 is the limiting case of the stable soliton solution when @xmath158 .
it seems as if stable soliton is asymptotically unstable
. however , there is no contradiction in this unusual situation .
this is because the stability analysis of sec .
[ sec4 ] deals with the infinitesimal perturbations , which preserve the monotonous character of the soliton solution . as a consequence ,
the scalar sector of the theory can be gauged away .
this is not the case with the trivial vacuum @xmath157 .
no matter how small the perturbations of this vacuum are , they can not be gauged away .
this is why the stability of @xmath157 can not be a substitute for the asymptotic stability of the soliton vacuum . in the second example
, i shall examine the scale factor of the form @xmath159 as seen from its graph in fig .
5 , it mimics the standard model of the universe . indeed , all the expected phases of the cosmological , height=139 ] evolution are there .
the inflationary epoch begins at infinite past , and lasts infinitely many @xmath138-folds .
the graceful exit from inflation happens at @xmath160 , when the early acceleration stops .
the universe continues to expand slowly , until it reaches the moment when the late time acceleration begins .
the present epoch is located at @xmath161 .
i shall demonstrate later how the parameter @xmath140 and the present time @xmath162 are calculated from the known values of the hubble and deceleration parameters .
the action functional whose vacuum solution is defined by ( [ 47 ] ) has the form ( [ 14 ] ) , with @xmath60 and @xmath46 calculated from ( [ 12 ] ) and ( [ 13 ] ) .
a straightforward procedure leads to the cumbersome expressions which i choose not to display here . instead , .[f6 ] , height=132 ]
their graphs are presented .
the action ( [ 14 ] ) , with @xmath60 and @xmath46 depicted in figs .
[ f6 ] and [ f7 ] , has the soliton solution @xmath64 .
it is always possible to simplify the action by .[f7 ] , height=132 ] a suitable redefinition of the scalar field .
for example , the redefinition @xmath163 of the form ( [ 43 ] ) makes the potential a periodic function of @xmath148 with the period @xmath153 . at the same time
, the target metric becomes a periodic function with the period @xmath164 .
this means that the new scalar @xmath148 lives on a circle .
the solution @xmath165 turns into the soliton solution shown in fig .
it represents the mapping of the real axis onto the half of the circle .
the parameter @xmath140 can be determined from the known values of the hubble and deceleration parameters .
these are defined as @xmath166 with the scale factor given by ( [ 47 ] ) , the parameters @xmath167 and @xmath168 depend on @xmath140 and @xmath169 only through the combination , height=151 ] @xmath170 .
the direct calculation yields the functions whose graphs are displayed in fig .
[ f8 ] . using these
, one easily finds that the astronomically observed values @xmath171
imply @xmath172 and @xmath173 .
thus , @xmath174 the obtained value of @xmath162 is the time coordinate measured from the arbitrarily chosen origin @xmath175 .
this is not what one would like to have . instead
, the present epoch should be measured relative to a physically significant moment in the history of the universe .
this can not be the big bang , as our model does not have one . instead , my choice is the end of inflation .
the end of inflation @xmath176 is naturally defined as the moment when the early acceleration stops .
it is seen from the graph in fig .
[ f5 ] that @xmath177 .
this leads to @xmath178 the time interval ( [ 51 ] ) is a substitute for what is commonly called the age of the universe .
the purpose of this work has been to demonstrate how an arbitrarily chosen universe dynamics can be made a solution of a simple model . to this end , i made use of the concept of geometric sigma models .
these models possess two distinctive features .
the first is that any metric can be made a solution of a particular geometric sigma model .
the second ensures that the complete matter content can be gauged away . in this paper ,
a geometric sigma model is associated with every homogeneous , isotropic and spatially flat geometry .
as it turns out , the model describes one scalar field in interaction with einstein s gravity .
it possesses the vacuum solution @xmath165 , @xmath179 @xmath180 @xmath181 .
technically , one would want the metric to be that of the standard model cosmology .
ultimately , the geometry of the universe is chosen to fit the astronomical observations .
the present work begins with the recapitulation of the concept of geometric sigma models .
the construction of the generic model is presented , and subsequently applied to the homogeneous , isotropic and spatially flat geometry of the universe .
then , the dynamics of small , localized perturbations of the vacuum is examined .
it is demonstrated how all but two degrees of freedom can be gauged away .
the classical stability of the gauge fixed linearized theory is proven in sec .
this is done by direct calculation , as stability of the vacuum solution is not guaranteed by the very construction of geometric sigma models .
the stability analysis concludes the general considerations of the paper . as a result
, a _ purely geometric sigma model is constructed , which accommodates a stable cosmological solution of whatever geometry is wanted_. the rest of the paper is devoted to examples .
the first is the example of an inflationary universe .
the second mimics the standard model cosmology .
the corresponding action functionals are constructed along the lines described in sec .
[ sec2 ] . the obtained target metric @xmath60 , and the potential @xmath46 are presented through their graphs .
the graphical method is also used for numerical calculations . in particular , it is demonstrated how the parameters of the model are calculated from the known values of the hubble and deceleration parameters . at the end , let me point out that the procedure described in this paper misses an important ingredient .
it concerns the presence of ordinary matter that every realistic model of the universe must take into account .
indeed , the only matter that geometric sigma models of sec .
[ sec2 ] deal with is a specific dark energy of purely geometric origin .
although it successfully generates any desirable geometry of the universe , the influence of ordinary matter is not taken into account .
the completion of the incomplete cosmology presented in this paper requires the inclusion of matter fields and their quantum fluctuations . until then
, it is comforting to know that dark energy is commonly believed to dominate all other forms of matter in the universe . owing to this
, the predictions of this work may not be far from realistic , after all . | geometric sigma model is a purely geometric theory in which spacetime coordinates are seen as scalar fields coupled to gravity .
although it looks like ordinary sigma model , it has the peculiarity that its complete matter content can be gauged away .
the remaining geometric theory possesses a vacuum solution that is predefined in the process of constructing the theory .
the fact that vacuum configuration is specified in advance is another peculiarity of geometric sigma models . in this paper ,
i construct geometric sigma models based on the arbitrarily chosen geometry of the universe .
whatever geometry is chosen , the dynamics of its small perturbations is shown to be classically stable . this way , any desirable metric is made a stable solution of a particular model .
the inflationary universe and the standard model universe are considered as examples of how this is done in practice . |
gaia observations of eclipsing binary stars will be of utmost importance to advances in stellar astrophysics . for
no other class of objects one could determine fundamental stellar parameters , i.e. absolute mass , size and surface temperature distribution with a comparable accuracy .
solutions of wide detached binaries can be used to accurately position them on the absolute h - r diagram .
identical age of both components places useful constraints on the theoretical isochrones for the given metallicity and rotational velocity which will also be derived from gaia observations .
components in short period systems are closer and mutually disturbed , so their evolution is different from that of single stars .
but accurate surface temperatures and sizes derived from binary solutions fix their luminosity and are useful to gauge their distance even for objects that are too far for the astrometric capabilities of the satellite ( see wythe & wilson 2002 )
. availability of on - board spectroscopy is vital to the study of eclipsing binaries .
semi - major axis and stellar masses could not be determined in any other way , and additional information on metallicity and rotational velocity helps in physical interpretation .
one might argue that this information could be obtained by ground - based follow - up observations . in our experience
this is not feasible . in asiago we launched an intensive campaign to spectroscopically observe eclipsing binaries discovered by hipparcos ( munari et al .
2001 [ hereafter m2001 ] , zwitter et al . 2003 ) . after three years we barely finished the spectroscopic coverage of the first 18 systems .
hipparcos discovered nearly 1000 systems , gaia will see hundred thousands .
these objects are distributed over the whole sky , so fiber optic spectroscopy can not reduce the required observing time significantly .
the strength of the gaia mission is in the numbers .
gaia will observe @xmath1 eclipsing binaries brighter than @xmath2 , @xmath3 of these will be double - lined systems ( m2001 ) . even if the stellar parameters will be determined at 1% accuracy only for 1% of them this
is still 25-times more than what has been obtained from all ground - based observations in the past ( cf .
andersen 1991 ) . moreover most of the gaia binaries will be of g - k spectral type ( cf .
zwitter & henden 2003 ) where there exists only a small number of systems with accurate solutions .
astrophysical importance of eclipsing binaries is discussed in other contributions ( milone 2003 , wilson 2003 , van hamme 2003 ) . here
we focus on our experience obtained from real stars that were observed in the gaia - like mode .
we start with discussion of how an object is recognized to be a binary and a determination of its orbital period .
next we discuss the accuracy of derivation of its fundamental parameters and the possibility to detect intrinsic variability of stars in binaries .
we close with some general remarks on the types of binaries that will be discovered .
we stress that huge numbers of objects call for completely automated reduction and possibly even interpretation techniques .
a large fraction of binary stars with orbital periods over a month that are closer than 1 kpc will be discovered astrometrically .
systems with periods of up to 10 years will be recognized due to their non - linear proper motion and those with periods of over a century will be resolved ( esa sp-2000 - 4 , arenou 2003 ) .
systems with orbital periods of less than a month will be mainly discovered by their photometric and spectroscopic variability .
gaia is unique because it will re - observe the same region of the sky many times over .
the number of transits for the spectroscopic focal plane will be around 100 with extremes a factor 2 higher or lower . the transits are not distributed evenly in time ( see fig . 1 ) .
this should pose no problems in analysis of binary stars if the satellite rotation and precession periods are kept incommensurable .
the sampling permits a good phase coverage of all orbital periods that are shorter than the mission lifetime .
also the duration of individual focal plane passages is just 100 seconds , so orbital motion smearing is negligible .
photometric variability does not need to be a consequence of the binarity of the source : pulsations , rotating and time - dependent stellar spots , as well as different types of semi - regular variables will be common among the g and k stars that will be the most frequent type of objects observed .
the best way to recognize that the detected photometric variability is indeed due to binarity is by establishing its repeatability and light curve shape ; so the orbital period needs to be determined . the same is true for spectroscopic observations .
the exceptions are of course double - lined binaries where a quarter - phase spectrum with well separated lines immediately points to the binary nature of the source
. potentials of photometry and spectroscopy for determination of orbital period are different , with spectroscopy being always preferable . as an example
let us examine a detached system gk dra ( fig . 2 ) .
it was discovered by the hipparcos satellite . but a search for orbital period from the 124 hipparcos observations proved unsuccessful .
there are several possible periods with the most likely solution of 16.96-days ( fig . 2 , top ) .
this value , which is also quoted in the hipparcos catalogue , is immediately shown to be wrong by only 35 spectroscopic observations in the gaia spectral window obtained by our gaia - like ground - based observing campaign ( zwitter et al .
2003 ; fig . 2 , middle ) .
the advantage of spectroscopic monitoring is in the fact that radial velocities are constantly changing with orbital phase .
so every point contributes to the orbital period search . in the case of photometry
the light curve out of eclipses is flat , so determination of orbital period is based only on a couple of points within the eclipses .
an extensive photometry can of course resolve this problem , but note how a periodogram from over 1300 photometric observations ( fig . 2 , bottom )
is still more ambiguous than the one obtained from only 35 spectroscopic observations in the gaia spectral window .
the strength of spectroscopy for the orbital period determination can also be seen from the light curves in figure 3 .
photometric information obtained by the gaia satellite will be far superior to that of hipparcos .
gaia will reach much fainter magnitudes and observe in many broad and narrow photometric bands .
but the number of epoch observations will be similar to that of hipparcos .
the coverage of gk dra to be obtained by gaia is given in figure 4 .
note that only a single observation in broad band filters falls within eclipses .
the eclipse coverage in intermediate passband filters is better .
broad- and narrow - band photometry can be used to measure orbital inclination as well as relative sizes and absolute temperatures of both stars .
but the orbital period itself will be much easier to determine from spectroscopy .
a total of 81 spectra are well distributed over orbital cycle , so the orbital period , semi - major axis , and both masses will be unambiguously measured from spectroscopic radial velocity measurements .
as mentioned in the introduction we are observing 18 hipparcos binaries in the gaia - like mode .
this means we are trying determine their orbital solution and fundamental parameters using only hipparcos ( @xmath4 , @xmath5 , @xmath6 ) photometry and ground based spectroscopy in the gaia spectral range .
spectroscopic data are extracted from a single echelle order observation with the 1.8-m telescope of the asiago observatory .
it turns out that such an approach is realistic as the accuracy of the solution is limited by a rather small ( @xmath7 ) number of hipparcos photometric measurements resulting in a poor coverage of eclipses
. this will be also the case with gaia .
the results we obtain should present a lower limit to the expected gaia accuracy , as we are using only rather noisy hipparcos photometry , while gaia photometry will have excellent precision and a much larger number of photometric bands .
figure 5 presents the light curve shapes of the overcontact binary v781 tau obtained in different narrow and broad passbands .
note that the photometric accuracy for objects brighter than @xmath8 will be better than 0.01 mag , so even rather subtle differences in the light curve shape and magnitude level for this binary with @xmath9 k will be easily discernible . in table 1
we quote accuracies of fundamental parameters for 6 systems published so far ( m2001 , zwitter et al .
we note that relative errors in most parameters are 2% or lower .
the exceptions are individual radii which are not well determined due to a scarce photometric coverage of eclipses
. solutions of uv leo and gk dra also have large uncertainties .
this is due to their intrinsic variability ( see below ) .
v781 tau is a binary which fills its roche lobe up to the l2 point .
temperatures and sizes of such binaries can be accurately determined .
so the distance can also be calculated with a remarkable accuracy .
many stars of g and k spectral types are intrinsic variables . this is true also for binary members . with this goal in mind
the observations in that gaia - like mode that use only hipparcos photometry with @xmath7 observations of each star were supplemented by devoted ground - based photometric campaigns ( dallaporta et al .
2000 , 2002 , 2002a , mikuz et al .
2002 , frigo et al .
2002 ) .
uv leo is a detached system with surface spots which cause a variation of system brightness by @xmath10 mag ( mikuz et al .
the system also showed a sudden change in the orbital period in feb .
1981 , possibly due to a passage of a low - mass third body .
devoted photometry of gk dra ( fig .
3 , bottom right ) shows unusually large scatter .
it turns out that the differences between the binary solution and observations are not due to noise but point to an intrinsic variability of @xmath11-sct type ( figure 6 ) .
a limited number of photometric and spectroscopic observations obtained by gaia will make it difficult to study intrinsic variability of the binary components .
still a large number of photometric bands will easily point to temperature changes that do not repeat with orbital cycle and are so due to intrinsic variability of the binary components .
interesting cases could be picked for detailed follow - up observations .
these include new interacting binaries ( cropper 2003 ) .
so far we discussed eclipsing double - lined spectroscopic binaries . in majority of cases
we will be less fortunate .
the systems could be too faint to obtain any useful spectroscopy , non - eclipsing or single - lined .
we discuss these in turn . for systems fainter than @xmath2 spectroscopic radial velocities will be difficult to measure even in double - lined cases .
these faint objects will far outnumber the bright spectroscopic binaries .
binarity will have to be established from eclipses or a reflection effect , both measured with a large number of photometric passbands but in a limited number of epochs .
some problems with the determination of orbital period of such systems have been discussed in section 2 .
here we only note that the binarity of these sources could be in general easily established due to a large photometric accuracy . in many cases the orbital ephemeris
could also be derived , therefore permitting ground - based spectroscopic follow - up observations at quarter phases establishing absolute masses and dimensions of the binary components . the easiest to recognize
will be systems close to contact , where much of the system information could be recovered from photometry alone .
eclipses will be rather uncommon , especially in the wide detached cases .
but for a double - lined non - eclipsing spectroscopic binary the mass ratio can still be easily calculated . and in not too wide binaries
the reflection effect can be measured from accurate photometry .
this constrains the temperatures and inclination of the system .
the inferred spectral classification can be finally checked against the properties of the spectra obtained close to quadratures where the spectral lines are well separated .
most binaries with the mass ratio below 0.3 will be single - lined , permitting to derive only a spectroscopic mass function .
gaia will discover huge numbers of spectroscopic and eclipsing binaries .
the numbers are orders of magnitude larger than everything collected in the last century from the ground . in many cases
the observations obtained by gaia will be good enough to determine system parameters at 1 - 2% accuracy level .
they will have an immense impact on theories of stellar structure and evolution . such a large data set requires an automation of all stages of reduction and interpretation .
no - one could recognize photometric eclipses or winging radial velocity curves in hundred - thousands or even millions of systems by eye .
but even interpretation and classification has to be completely automatic with only the most unusual cases to be marked for human inspection .
wythe & wilson ( 2001 , 2002 ) successfully classified some photometric eclipsing binaries from the ogle database with semi - automatic procedures .
prsa ( 2003 ) obtained some encouraging results for double - lined eclipsing binaries .
clearly development of reliable classification and analysis procedures is one of the major tasks facing the scientific community before the launch of gaia . + * acknowledgement .
* generous allocation of observing time by osservatorio astronomico di padova and financial support from the ministry of education , science and sport of slovenia are acknowledged .
andersen j. 1991 , a&arv , 3 , 91 arenou f. 2003 , this volume cropper m. 2003 , this volume dallaporta s. , tomov t. zwitter t. , munari u. 2000 , ibvs , 4990 , 1 dallaporta s. , tomov t. , zwitter t. , munari u. 2002 , ibvs , 5312 , 1 dallaporta s. , tomov t. , zwitter t. , munari u. 2002a , ibvs , 5319 , 1 frigo a. , piccoli p. , setti a. , tomasoni s. , tomov t. , munari u. , marrese p.m. , zwitter t. 2002 , ibvs , 5359 , 1 mikuz h. , dintinjana b. , prsa a. , munari u. , zwitter t. 2002 , ibvs , 5338 , 1 milone e.f . 2003 , this volume munari u. 1999 , balt .
a. , 8 , 73 munari u. , tomov t. , zwitter t. , milone e.f .
, kallrath j. , marrese p.m. , boschi f. , prsa a. , tomasella l. , moro d. 2001 , aa , 378 , 477 [ m2001 ] prsa a. 2003 , this volume stellingwerf r.f .
1978 , apj , 224 , 953 van hamme w. 2003 , this volume wythe j.s.b . , wilson r.e .
2001 , apj , 559 , 260 wythe j.s.b . , wilson r.e .
2002 , apj , 571 , 293 wilson r.e .
2003 , this volume zwitter t. , munari u. , marrese p.m. , prsa a. , milone e.f .
, boschi f. , tomov t. 2003 , aa , submitted zwitter t. , henden a. 2003 , this volume | gaia observations of eclipsing binary stars will have a large impact on stellar astrophysics .
accurate parameters , including absolute masses and sizes will be derived for @xmath0 systems , orders of magnitude more than what has ever been done from the ground .
observations of 18 real systems in the gaia - like mode as well as with devoted ground - based campaigns are used to assess binary recognition techniques , orbital period determination , accuracy of derived fundamental parameters and the need to automate the whole reduction and interpretation process .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
we are grateful to the open science grid for the computing resources .
the author thanks dirk rischke for providing the 1 fluid hydrodynamics code . h.p . acknowledges a feodor lynen fellowship of the alexander von humboldt foundation . this work was supported in part by u.s .
department of energy grant de - fg02 - 05er41367 .
thanks jan steinheimer for help with the extension of the equation of state , guangyou qin for providing the glauber calculation of the numebr of particpants and steffen a. bass and berndt mller for fruitful discussions .
99 k. aamodt _ et al . _
[ alice collaboration ] , phys .
lett . * 106 * , 032301 ( 2011 ) [ arxiv:1012.1657 [ nucl - ex ] ] .
k. aamodt _ et al . _
[ alice collaboration ] , phys .
b * 696 * , 30 ( 2011 ) [ arxiv:1012.1004 [ nucl - ex ] ] .
k. aamodt _ et al . _
[ the alice collaboration ] , arxiv:1011.3914 [ nucl - ex ] .
k. aamodt _ et al . _
[ the alice collaboration ] , phys .
lett . * 105 * , 252301 ( 2010 ) [ arxiv:1011.3916 [ nucl - ex ] ] .
n. armesto _ et al .
_ , j. phys .
g * 35 * , 054001 ( 2008 ) [ arxiv:0711.0974 [ hep - ph ] ] .
m. luzum , phys .
c * 83 * , 044911 ( 2011 ) [ arxiv:1011.5173 [ nucl - th ] ] .
h. niemi , g. s. denicol , p. huovinen , e. molnar and d. h. rischke , arxiv:1101.2442 [ nucl - th ] . h. song , s. a. bass and u. w. heinz , arxiv:1103.2380 [ nucl - th ] .
r. a. lacey , a. taranenko , n. n. ajitanand and j. m. alexander , phys .
c * 83 * , 031901 ( 2011 ) [ arxiv:1011.6328 [ nucl - ex ] ] .
s. a. bass and a. dumitru , phys .
c * 61 * , 064909 ( 2000 ) [ arxiv : nucl - th/0001033 ] .
d. teaney , j. lauret and e. v. shuryak , arxiv : nucl - th/0110037 .
t. hirano , u. w. heinz , d. kharzeev , r. lacey and y. nara , phys .
b * 636 * , 299 ( 2006 ) [ arxiv : nucl - th/0511046 ] . c. nonaka and s. a. bass , phys .
c * 75 * , 014902 ( 2007 ) [ arxiv : nucl - th/0607018 ] . t. hirano , p. huovinen and y. nara , arxiv:1012.3955 [ nucl - th ] . b. schenke , s. jeon and c. gale , arxiv:1102.0575 [ hep - ph ] .
r. andrade , f. grassi , y. hama , t. kodama and o. .
j. socolowski , phys .
lett . *
97 * , 202302 ( 2006 ) [ arxiv : nucl - th/0608067 ] .
b. m. tavares , h. j. drescher and t. kodama , braz .
j. phys .
* 37 * , 41 ( 2007 ) [ arxiv : hep - ph/0702224 ] .
h. petersen , j. steinheimer , g. burau , m. bleicher and h. stocker , phys .
c * 78 * , 044901 ( 2008 ) [ arxiv:0806.1695 [ nucl - th ] ]
. h. holopainen , h. niemi and k. j. eskola , phys .
c * 83 * , 034901 ( 2011 ) [ arxiv:1007.0368 [ hep - ph ] ] .
k. werner , i. karpenko , t. pierog , m. bleicher and k. mikhailov , phys . rev .
c * 82 * , 044904 ( 2010 ) [ arxiv:1004.0805 [ nucl - th ] ] .
b. schenke , s. jeon and c. gale , phys .
lett . * 106 * , 042301 ( 2011 ) [ arxiv:1009.3244 [ hep - ph ] ] .
t. renk , h. holopainen , r. paatelainen and k. j. eskola , arxiv:1103.5308 [ hep - ph ] .
h. holopainen , s. rasanen and k. j. eskola , arxiv:1104.5371 [ hep - ph ] .
h. petersen , t. renk and s. a. bass , phys .
c * 83 * , 014916 ( 2011 ) [ arxiv:1008.3846 [ nucl - th ] ] .
h. petersen , g. y. qin , s. a. bass and b. muller , phys .
c * 82 * , 041901 ( 2010 ) [ arxiv:1008.0625 [ nucl - th ] ] .
h. petersen , c. coleman - smith , s. a. bass and r. wolpert , j. phys .
g * 38 * , 045102 ( 2011 ) [ arxiv:1012.4629 [ nucl - th ] ] .
s. a. bass _ et al .
. part .
* 41 * , 255 ( 1998 ) [ prog .
phys . * 41 * , 225 ( 1998 ) ] [ arxiv : nucl - th/9803035 ] .
m. bleicher _ et al .
_ , j. phys .
g * 25 * , 1859 ( 1999 ) [ arxiv : hep - ph/9909407 ] .
h. petersen , m. bleicher , s. a. bass and h. stocker , arxiv:0805.0567 [ hep - ph ] .
b. nilsson - almqvist and e. stenlund , comput .
commun . * 43 * , 387 ( 1987 ) .
t. sjostrand , comput .
commun .
* 82 * , 74 ( 1994 ) .
d. h. rischke , s. bernard and j. a. maruhn , nucl .
a * 595 * , 346 ( 1995 ) [ arxiv : nucl - th/9504018 ] .
d. h. rischke , y. pursun and j. a. maruhn , nucl .
a * 595 * , 383 ( 1995 ) [ erratum - ibid .
a * 596 * , 717 ( 1996 ) ] [ arxiv : nucl - th/9504021 ] .
j. steinheimer , v. dexheimer , h. petersen , m. bleicher , s. schramm and h. stoecker , phys .
c * 81 * , 044913 ( 2010 ) [ arxiv:0905.3099 [ hep - ph ] ] .
j. steinheimer , s. schramm and h. stocker , arxiv:0909.4421 [ hep - ph ] .
h. petersen , j. steinheimer , m. bleicher and h. stocker , j. phys .
g * 36 * , 055104 ( 2009 ) [ arxiv:0902.4866 [ nucl - th ] ] .
m. mitrovski , t. schuster , g. graf , h. petersen and m. bleicher , phys .
c * 79 * , 044901 ( 2009 ) [ arxiv:0812.2041 [ hep - ph ] ] . | the first results from heavy ion collisions at the large hadron collider for charged particle spectra and elliptic flow are compared to an event - by - event hybrid approach with an ideal hydrodynamic expansion .
this approach has been shown to successfully describe bulk observables at rhic . without changing any parameters of the calculation the same approach
is applied to pb+pb collisions at @xmath0 tev .
this is an important test if the established understanding of the dynamics of relativistic heavy ion collisions is also applicable at even higher energies .
specifically , we employ the hybrid approach with two different equations of state and the pure hadronic transport approach to indicate sensitivities to finite viscosity .
the centrality dependence of the charged hadron multiplicity , @xmath1 spectra and differential elliptic flow are shown to be in reasonable agreement with the alice data . furthermore , we make predictions for the transverse mass spectra of identified particles and triangular flow .
the eccentricities and their fluctuations are found to be surprisingly similar to the ones at lower energies and therefore also the triangular flow results are very similar .
any deviations from these predictions will indicate the need for new physics mechanisms responsible for the dynamics of heavy ion collisions .
recently , the first results from heavy ion collisions at @xmath2 tev have been published by the alice collaboration @xcite . to study strongly interacting matter at high temperatures
has been the goal of the relativistic heavy ion program at the relativistic heavy ion collider ( rhic ) since more than a decade .
the 10 times higher beam energies at the large hadron collider ( lhc ) allow for the investigation of the dynamical evolution of nucleus - nucleus collisions that have been established in au+au collisions at @xmath3 gev in a different kinematic range @xcite .
it is especially interesting , if the matter still behaves as a almost perfect liquid or if the quark gluon plasma becomes more viscous by going to higher temperatures @xcite .
hybrid approaches that are based on hydrodynamics for the hot and dense stage of the evolution that is coupled to hadronic transport approaches to describe the successive decoupling of the matter have been very successful in describing the properties of the bulk matter that is created at rhic @xcite and lhc @xcite . within the last year , full event - by - event hydrodynamic approaches
@xcite have become more favorable because it has turned out that the effect of initial state fluctuations on final flow observables needs to be studied in a consistent way to draw quantitative conclusions for e.g the value of the shear viscosity .
it is important to apply well - established approaches for the dynamical evolution of heavy ion reactions at rhic to the higher energy collisions at lhc without tuning parameters to investigate how good the energy dependence is described by a specific model .
differences between the experimental data and the theoretical calculations imply the need for new physics concepts to be applied . to understand the bulk evolution at lhc energies is a pre - requisite for all further detailed studies of e.g. jet quenching and energy loss or electromagnetic probes @xcite . in this manuscript
a event - by - event hybrid approach based on the ultra - relativistic quantum molecular dynamics ( urqmd ) transport approach with an embedded ( 3 + 1 ) dimensional ideal hydrodynamic evolution is applied to lead - lead collisions at lhc energies .
first , a comparison to the available experimental data on charged particle multiplicities , @xmath4 spectra and elliptic flow is carried out with the exact same parameter set that has been applied at rhic energies .
then , transverse mass spectra and triangular flow for identified particles are predicted . to compare the amount of initial state fluctuations , the probability distributions of coordinate space eccentricities
are compared to their corresponding values at rhic .
let us start with a short description of the hybrid approach that has been developed for sps energies @xcite and recently successfully applied to gold - gold collisions at the highest rhic energy @xcite .
the early non - equilibrium evolution is described by the urqmd approach @xcite , where the two lead nuclei are initialized according to wood - saxon profiles followed by binary interactions of the nucleons .
the main contribution to the particle production at high energies is achieved by string excitation and fragmentation where for the hard collisions ( momentum transfer @xmath5 gev ) are treated by pythia @xcite . at the so called starting time of @xmath6 fm
the particle distributions are transferred to energy , momentum and net baryon density distributions by representing each particle with a three - dimensional gaussian distribution ( width @xmath7 fm ) that is lorentz - contracted along the beam direction . during the following ideal relativistic one fluid evolution @xcite two different equations of state are employed , one representing a hadron gas ( hg - eos ) and one including a cross - over deconfinement phase transition based on a chiral approach ( de - eos)@xcite .
the transition back to the hadronic transport approach happens on a constant proper time hypersurface , where the cooper - frye equation is applied on transverse slices of thickness @xmath8 fm that have cooled down below an energy density of @xmath9 mev/@xmath10 @xcite .
this approach provides the full final phase space distributions of the produced particles for each event and can be compared to the pure transport approach by turning off the hydrodynamic evolution which allows for a qualitative study of viscous effects .
the first observable to look at is the charged particle multiplicity at midrapidity . in fig .
[ fig_mulimpch ] the calculation of the centrality dependent multiplicity scaled by the number of participants ( estimated in a glauber approach ) is shown .
the hadronic transport approach urqmd provides a reasonable description of the multiplicity .
for central collisions the predictions published in @xcite are right on top of the alice data while with decreasing centrality the number of charged particles is a little lower than in the data .
this fair agreement with the data hints to the fact that the main particle production can be described by the initial binary nucleon - nucleon interactions treated by pythia .
the hydrodynamic evolution does not affect the particle production .
since ideal hydrodynamics implies an isentropic expansion this means that the charged particle multiplicity is determined in the initial state and by the final resonance decays .
for the following calculations of spectra and collective flow four different centrality classes have been chosen that match the ones applied by the alice collaboration as they are listed in the following table : [ cols="^,^",options="header " , ] the transverse momentum spectrum for charged particles in the mentioned centrality classes are compared to experimental data in the most central bin ( see fig .
[ fig_dndptimpch ] ) .
the main difference between the hybrid and the transport calculation is in the slopes of the spectra .
as expected the hydrodynamic evolution leads to a purely exponential @xmath1 dependence which describes the data until @xmath11 gev very well . at higher transverse
momenta the power law tail from hard processes becomes important for a good agreement with the measured values . in the range from 4 to 6 gev
the non - equilibrium description exemplified by the urqmd calculation provides a better description of the experimental data . in fig .
[ fig_dndmtimpall ] predictions for the transverse mass spectra at midrapidity of pions , kaons and protons are presented .
the pion spectra are very similar to the charged particle spectra since they represent the major fraction of the newly produced particles in the collision .
kaons are strange mesons and protons are chosen because they have a higher mass and are baryonic degrees of freedom .
the general features of the transverse mass spectra are similar to the ones observed at rhic and imply a collective radial velocity that drives all the particle species .
the two different equations of state lead to very similar results with the deconfinement transition having a little steeper slope due to the more rapid expansion due to the higher pressure in the quark gluon plasma phase . after proving a rather successful agreement with basic quantities like the multiplicity and transverse momentum spectrum
the next step is to look at anisotropic flow observables .
the elliptic flow has been calculated with respect to the reaction plane by averaging over all charged particles in all events to be compared to the alice measurement that relies on the four - particle cumulant method in two centrality bins .
fig .
[ fig_v2ptimpch ] shows a good agreement between the hybrid calculations and the data , especially between @xmath1=0.8 - 2.5 gev . in the very low transverse momentum region
the hybrid approach underpredicts the data which has been observed in other calculations as well @xcite . at higher
@xmath1 again the influence of hard processes needs to be taken into account . to quantify the shape of the initial conditions employed for the hydrodynamic calculation and its event - by - event fluctuations fig .
[ fig_probeps ] shows the probability distribution of the coordinate space asymmetry characterized by the eccentricity and the triangularity as defined in @xcite .
the initial @xmath12 coefficients have been calculated in each event and the normalized probability distribution is plotted for two different centrality bins . for central collisions
the mean value and the shape of the distributions are very similar for the participant eccentricity and the triangularity since both of them are mainly generated by fluctuations . for more peripheral collisions
the eccentricity is influenced by a large geometry component due to the ellipsoidal shape of the initial state in the transverse plane .
therefore , the mean eccentricity is larger and the fluctuations increase leading to a wider distribution , whereas the triangularity stays smaller and the distribution has a smaller width .
since the triangularity has been introduced because of its sensitivity to initial state fluctuations the higher multiplicity at lhc energies triggers the expectations that the fluctuations become smaller compared to rhic energies . in fig .
[ fig_probeps ] the triangles and diamonds depict the eccentricity and triangularity calculation from urqmd initial conditions for au+au collisions at @xmath3 gev .
surprisingly , the @xmath12 distributions match almost exactly the ones at lhc energies for the two similar centrality classes . to calculate the transverse momentum dependence of the anisotropic flow coefficients for identified particles ( see fig .
[ fig_v3ptid ] ) the event plane method including removal of auto - correlations and resolution correction has been applied as described in more detail in @xcite . in this way
, the predictions can be directly compared to a future experimental result for minimum bias pb+pb collisions at @xmath2 tev .
both elliptic and triangular flow show the expected mass splitting between pions and protons .
furthermore , the elliptic flow has roughly double the value of the triangular flow result as it has previously been calculated for rhic energies .
the anisotropic flow are not very sensitive to the choice of the equation of state which points to the fact that the interplay of freeze - out transition and flow development due to pressure gradients results in very similar values independent of the existence of a quark gluon plasma phase . finally in fig .
[ fig_v3ptch ] predictions for the transverse momentum dependence of triangular flow for charged particles in the four different centrality classes are made .
there is a weak centrality dependence , but overall the different curves from the hybrid calculation look very similar qualitatively . since the calculation with the deconfinement phase transition leads to very similar results , only the hadron gas result is shown in this figure to reduce the number of lines displayed . to investigate the effect of a finite viscosity during the expansion the calculation
is compared to the pure urqmd result . within the very viscous hadronic transport approach
the pressure gradients are too small to transfer the initial coordinate space asymmetry to a final state momentum space asymmetry . for impact parameters up to @xmath13 fm
the result is consistent with zero and even for very peripheral event the result is negligible especially in the low transverse momentum region where most of the particles are .
this leads to the conclusion that a finite triangular flow measurement is a strong indication for almost ideal hydrodynamic behavior during the hot and dense evolution of the heavy ion reaction . to summarize
we have presented a comparison of bulk observables measured in pb+pb collisions at @xmath2 tev to a state - of - the - art event - by - event hybrid description .
employing the same parameters as have been used for au+au collisions at rhic a reasonably good agreement for the multiplicity , transverse momentum spectra and elliptic flow is achieved .
one can conclude that the basic assumptions about the major ingredients that are needed to describe the dynamic evolution of heavy ion reactions do not have to be revised at lhc energies . with a non - equilibrium initial state , an ideal hydrodynamic evolution and a hadronic afterburner predictions for identified particle transverse mass spectra and elliptic as well as triangular flow
are made .
the initial state fluctuations quantified by the probability distribution of the initial eccentricity and triangularity are very similar at lhc and rhic energies within the urqmd approach . a large viscosity during the evolution results in almost negligible higher harmonic coefficients ,
so the triangular flow measurement at lhc will provide a robust confirmation of the almost ideal hydrodynamic expansion . |
the precise determination of light quark masses , @xmath5 , from full qcd simulations is of considerable interest since the lattice provides the only known access to their absolute scale .
we have pointed out some time ago @xcite that vacuum polarization effects have sizeable impact on extracting @xmath5 from the empirical @xmath6 ratio .
this became manifest in a substantial ambiguity in the determination of @xmath5 from the vector ward identities .
indeed , there is a freedom of chiral extrapolation ( @xmath0e ) on the lattice : at @xmath7 _
e.g. _ , the extrapolation along the line of equal valence and sea quark hopping parameters , @xmath8 ( direct extrapolation ) , yields a value for @xmath9 which is roughly a factor of two below the result from a semi - quenched , two - step extrapolation procedure . in the latter one first determines the values of @xmath10 for each ensemble with fixed value of @xmath11 ( similar to quenched simulations ) . for each @xmath12 @xmath13
is computed and , subsequently , this function is extrapolated to the physical point @xmath14 , at this given value of @xmath15 .
furthermore it has been shown by cp - pacs that the ambiguity seen at finite @xmath16 goes away after ce @xcite .
cp - pacs has used an improved action that allowed them to go to rather coarse lattices , providing them with a long lever arm for the various ce . in the present context ,
their lever arm turned out to be of little value , however , since the extrapolation with a linear ce led them to a very poor @xmath17 with such an ansatz , invalidating the small error on their value of @xmath5 .
given this situation we are motivated to resume the question with the standard wilson fermionic action and attempt a scaling analysis , by complementing previous sesam / t@xmath0l data with another set of vacuum configurations at @xmath18 , as specified in table [ tab : params ] . [
cols="^,^,^,^,^,<,<,<,<",options="header " , ]
the determination of light quark masses can be sketched by the following steps : + ( a ) fix @xmath19 from the @xmath0e of @xmath20 ; + ( b ) get @xmath21 from @xmath22 using the @xmath0e of @xmath23 ; + ( c ) determine @xmath24 via @xmath25 ; + ( d ) compute the quark mass in the lattice - scheme : @xmath26 ( e ) convert into @xmath27-scheme using @xmath28 ( f ) rescale the quark mass to @xmath29 ; + ( g ) extrapolate @xmath30 .
@xmath31 within this work was determined using a tadpole improved perturbative procedure @xcite . in the case of the above - quoted semi - quenched @xmath0e procedure the results for @xmath32 have to be extrapolated to the physical mass of the sea quarks before sending @xmath30 .
in order to achieve appropriate statistical errors throughout this analysis one has to take into account the correlation between the observables .
these correlations occur especially within the semi - quenched analysis , since all observables inside a @xmath0e are determined on the same set of configurations .
therefore all errors are determined using the jackknife procedure within all steps ( a ) to ( g ) sketched above .
[ fig : mlight_b5.6 ] shows the results for @xmath5 in the @xmath27-scheme for @xmath3 versus @xmath11 .
obviously the semi - quenched values for @xmath33 on the small lattice and for @xmath34 on the larger lattice deviate from the general behavior of the data .
a more detailed investigation shows that this is due to finite - size effects .
hence these two points are neglected in the extrapolation to the physical sea quark mass .
furthermore we reconfirm the above mentioned discrepancy between the results for the semi - quenched and the direct extrapolations .
[ fig : mlight_b5.5 ] displays the corresponding results for @xmath4 . here
the value for @xmath35 clearly shows finite - size effects and is hence discarded . in @xmath36 . ]
[ fig : mlight_vs_a ] presents the ce @xmath37 , with @xmath16 as obtained from @xmath38 at @xmath21 .
the continuum value for the light quark mass is determined using a constrained fit to the semi - quenched and direct values at two lattice constants each .
the fit is also shown in the figure .
as a final result for the light quark mass we get @xmath39 within the bigger errors of our analysis , this number is in good agreement with the result quoted by cp - pacs @xcite .
in contrast to the light quark mass there is no reasonable way to carry out a direct @xmath0e for the mass of the strange quark .
this is due to the fact that directly extrapolated results would describe mesons with strange valence quarks within a sea of strange sea quarks .
indeed , nature is given by a sea of light sea quarks and therefore only the semi - quenched @xmath0e makes sense . on the other hand
there are several ways to determine @xmath41 , the hopping - parameter corresponding to the mass of the strange quark . in this analysis we investigated three different inputs : @xmath42 , @xmath43 and @xmath44 . from a semi - quenched analysis of @xmath40 , one obtains the results displayed in fig .
[ fig : mstrange_vs_a ] .
the data clearly show deviations for the different definitions of the quark mass which are expected to disappear in the continuum .
hence again a constrained fit was done for the ce . in @xmath36 . ]
the result of this fit for the mass of the strange quark is @xmath45 obviously the strange quark mass needs all the more a larger range of @xmath16-values for a safe ce .
we conclude that ( _ i _ ) statistical errors of 5 to 10 % can be achieved before ce , ( _ ii _ ) with additional simulations at smaller @xmath15 , a reliable @xmath5-determination with 10 % accuracy is within reach .
we thank g. bali for computing the @xmath38 data , with support of the ec under contract hprn - ct-2000 - 00145 hadrons / lattice qcd . n.e .
thanks desy for partly supporting his phd project .
b.o . was supported by the dfg graduiertenkolleg `` feldtheoretische methoden in der elementarteilchentheorie und statistischen physik '' .
9 n. eicker , _ et al .
_ , phys . lett .
* b407 * ( 1997 ) 290 .
a. ali khan _ et al .
_ , hep - lat/0105015 ( 2001 ) .
b. orth , _ et al .
_ , these proceedings .
b. bolder _ et al .
_ , phys . rev .
* d62 * ( 2000 ) 054503 .
p. lepage _ et al .
_ , phys . rev . *
d 48 * ( 1993 ) 2250 . | we present new data on the mass of the light and strange quarks from sesam / t@xmath0l .
the results were obtained on lattice - volumes of @xmath1 and @xmath2 points , with the possibility to investigate finite - size effects .
since the sesam / t@xmath0l ensembles at @xmath3 have been complemented by configurations with @xmath4 , moreover , we are now able to attempt the continuum extrapolation ( ce ) of the quark masses with standard wilson fermions . |
the steady state solution @xmath21 of ( [ e : master ] ) is found as ( setting @xmath22 from now on , for convenience ) @xmath23 \label{e : st3}\ ] ] where @xmath24 , and @xmath25 , the asymptotic jump rate , is fixed by normalization of @xmath26 .
it is straightforward to show from ( [ e : st3 ] ) that in the limit of slow flows @xmath27 , the steady state stress response @xmath28 ( in an obvious notation ) is always linear : @xmath29 .
hence there is no yield stress for any choice of @xmath30 .
this contrasts with a model having an exponential distribution of barrier heights @xmath11 for different elements , which does show onset of a yield stress , connected with the presence of a glass transition , as @xmath17 is reduced @xcite . for monodisperse @xmath11 , as here , there is no such transition .
we now show that the steady state flow curve , for _ any _ choice of function @xmath20 , has a monotonically increasing @xmath31 .
first , by differentiating @xmath32 and using ( [ e : st3 ] ) , we find @xmath33 similarly , the normalisation integral for @xmath26 can be differentiated with respect to @xmath7 to give @xmath34 combining these two expressions ( with @xmath35 ) produces @xmath36 which is positive since @xmath37 is monotone increasing .
thus there are no regions on the flow curve with negative slope , and for no choice of @xmath20 does one expect any shear - banding instability to arise . however , the argument does not prove that a steady state is ever actually achieved .
we have found that it is reached under conditions of imposed strain @xmath38 , but _ not _ under an imposed stress @xmath3 ( for appropiate @xmath20 ) , for which the model exhibits temporal oscillations instead .
this numerical finding ( detailed below ) is also supported by a linear stability analysis for perturbations about the steady state flow , described fully in @xcite .
such an analysis shows that the transient behaviour close to steady state is never a real exponential decay but always has an oscillatory component .
when the decay rate of these oscillations changes sign , the flow becomes unstable to permanent oscillation .
the time evolution of the system at fixed stress was studied numerically by methods described in @xcite .
( the problem is not standard and significant amounts of computer time were required to achieve reliable results . )
a range of different functions @xmath20 and of @xmath11 values were tried .
varying @xmath3 over a wide range of values in each case reveals that steady flow is always reached for sufficiently small and sufficiently large imposed stresses ; but for intermediate @xmath3 , an oscillatory regime is observed whenever @xmath20 is strongly enough decreasing .
it is impractical to explore the entire function space for @xmath20 so at present the precise criteria for the presence of oscillations are not known . within the oscillatory regime
, we have so far found only single - period oscillations with no sign of period doubling or other more complex nonsteady behaviour that might point to the onset of rheo - chaos in some parts of the parameter space .
some examples of the oscillatory behaviour are given in fig .
[ f : osc_nonmon_eg](a ) . to obtain these results we took @xmath39 for @xmath40 and @xmath41 for @xmath42
; in addition we set @xmath43 .
the strain @xmath38 is shown for three different imposed @xmath3 . on close inspection
the oscillations , after a brief transient , appear perfectly periodic .
this applies not only in the stress but in the underlying distribution @xmath5 , which determines the future evolution .
hence there seems little doubt that these oscillations will persist indefinitely . from fig .
[ f : osc_nonmon_eg](a ) it is remarkable that the _ mean _ strain rate @xmath44 , defined as @xmath7 averaged over a single period of oscillation , is clearly a _ decreasing _ function of @xmath3 , in complete contrast to the monotonic , steady state , flow curve .
the time averaged values @xmath44 as read off from the simulations are plotted against @xmath3 in fig .
[ f : osc_flowcurve](b ) , overlayed with the steady state flow curve . within the oscillatory regime ,
the @xmath44 line deviates from the flow curve to the extent that it loses monotonicity over a large interval of @xmath3 .
if the oscillations were not detected but only the average behaviour measured , this would resemble the flow curve of a discontinuously shear - thickening system close to a jamming transition , although in that case shear banding could be expected as well .
( such a flow curve is indeed found in models where @xmath17 depends not on the local stress @xmath1 but only on its global mean , @xmath3 ; see @xcite . )
upper and lower transition points between oscillatory and steady flow can be seen in the figure . at each transition point
@xmath45 , the period of oscillation remains finite whilst the amplitude vanishes smoothly according to @xmath46 with @xmath47 .
as we discuss in our longer work @xcite , technical difficulties have prevented us from reliably fixing the asymptotic values of @xmath48 .
in particular we have been unable to rule out @xmath49 for both transitions , as expected for a hopf bifurcation @xcite ( although since the control parameter is an integral constraint , it is not clear to us if it should be a hopf bifurcation ) .
though always perfectly periodic , the waveform of the oscillations varies with @xmath3 .
close to either transition , the oscillations are near - sinusoidal , as demonstrated in fig .
[ f : osc_sine](a ) . further into the oscillatory regime , @xmath7
can no longer be decomposed into a single harmonic , but instead approaches a waveform in which most of the variation in @xmath7 is compressed into a small fraction of the total period of oscillation .
an example is given in fig .
[ f : osc_sine](b ) .
this behaviour at first appears to be similar to the ` stick slip ' motion observed in systems such as earthquakes @xcite , ultra - thin liquid films @xcite and granular media @xcite .
however , the underlying physics in our model seems to be somewhat different to these examples .
indeed , stick slip is usually viewed as a surface phenomenon , whereas the model studied in this paper has no surface and describes only bulk effects .
irrespective of the waveform , the product of @xmath44 and the period of oscillation @xmath50 is approximately constant , @xmath51 for this example .
this will be explained below in terms of the evolution of @xmath5 .
snapshots of @xmath5 during a single period of oscillation are given in fig .
[ f : snapshots_mono ] . the mechanism behind the oscillations can be qualitatively described in terms of two coexisting populations of elements a ` hot ' population of unstrained or slightly strained elements with @xmath52 , and hence a high effective temperature @xmath20 ; and a ` cold ' population of elements with @xmath53 and therefore a low @xmath20 .
starting at a time when @xmath54 is near to is minimum value , corresponding to the first snapshot in the figure , the stress weighted yield rate @xmath55 for both populations of elements is low .
this is because the hot elements have small @xmath1 , whereas the cold elements have a low yield rate @xmath0 .
thus the strain rate @xmath7 required to maintain the imposed constant mean stress @xmath3 , which is @xmath56 ( as seen by multiplying ( [ e : master ] ) by @xmath57 and integrating ) , is also low .
although @xmath54 is small in this state , it is nonetheless non zero , and therefore the strains @xmath1 of the cold elements increase according to @xmath58 .
this decreases their effective energy barrier @xmath59 , increasing their hop rate and therefore the rate of stress lost due to to cold elements yielding .
thus @xmath60 will also increase , which accelerates the rate at which the cold elements yield , and so on .
this description is that of a _ positive feedback loop _ that causes @xmath7 to increase at ever faster rates until all of the cold elements have been depleted . the second snapshot in fig .
[ f : snapshots_mono ] show the state of the system shortly before this has happened .
while @xmath54 is high , the hot elements are becoming rapidly strained and have little time to yield before crossing over into the cold region with large @xmath1 .
thus @xmath5 is flat for small @xmath1 .
as @xmath54 again decreases , this flat part of the distribution will start to decay as the elements within it yield .
however , the yield rate depends on @xmath20 , and since this changes to a lower value at @xmath61 , @xmath5 will decay more rapidly for @xmath1 smaller than @xmath62 than for @xmath1 greater than @xmath62 .
thus a dip will occur around the point @xmath63 , which can clearly be seen in fig .
[ f : snapshots_mono ] .
this dip becomes more pronounced with time until the system can again be viewed as coexisting ` hot ' and ` cold ' populations , when the cycle begins again .
a possible interpretation of the parameter @xmath64 , defined previously , is that it is the amount by which the system needs to be strained until the positive feedback loop just described starts to dominate the system behaviour , causing it to ` reset ' to the start of its cycle .
if this is correct , then @xmath64 should correspond to the point at which highly strained elements have the same yield rate as unstrained elements , _
i.e. _ @xmath65 .
rough calculations based on this assumption give @xmath66}$ ] , which for this example predicts @xmath67 , in fair agreement with the observed value .
within a simple model of shear thickening , we have found regimes where steady flow at constant imposed stress is unstable and temporal oscillation occurs instead .
known instances of such rheological instability have often been explained in terms of the spatial coexistence of subpopulations , or phases .
for instance , the temporal oscillations in viscosity observed in wormlike micelles under an imposed stress was attributed to a slowly fluctuating interface between a fluid phase and shear - induced structures @xcite . for surfactant solutions in the lyotropic lamellar phase ,
it was attributed to coexisting ordered and disordered phases @xcite . by contrast
, the temporal oscillations observed in our model arise even though they are by assumption spatially homogeneous , and the flow curves are everywhere monotonic .
it may be that , in practice , such rheo - oscillations will always be coupled to spatial heterogeneity of some sort .
experiments are not advanced enough to indicate whether this is the case .
one distinguishing feature of our equations is the role of memory effects , rather than inertia , in allowing oscillations . in our model ,
the memory resides in @xmath5 which can have different shapes for the same applied stress @xmath68 .
there may thus be a link to work on ` delay differential equations ' in which a first order differential equation for a state variable contains a feedback term that depends on the same state variable at an earlier time @xcite .
such equations are capable of showing a range of instabilities , including chaos , despite the absence of inertia .
the nature of this link remains a topic of current research .
its exploration might help answer the following question : what features , if any , could be added to our simple rheological model of shear thickening to allow rheo - chaos to arise _ without _ coupling to spatial degrees of freedom @xcite .
this could in turn help answer another important question : what is the dimension of the strange attractors that arise in rheo - chaos @xcite ?
in conclusion we suggest that the study of temporal ( or spatiotemporal ) pattern formation in non - newtonian flows ( at effectively zero reynolds number ) will form a major topic of experimental and theoretical research in the coming years . _
acknowledgements : _ the authors would like to thank suzanne fielding for stimulating discussions concerning this work .
aa also wishes to thank the university of edinburgh for the hospitality that allowed this work to start .
dah was funded by epsrc(uk ) grant no .
gr / m09674 . | we study the strain response to steady imposed stress in a spatially homogeneous , scalar model for shear thickening , in which the local rate of yielding @xmath0 of mesoscopic ` elastic elements ' is not monotonic in the local strain @xmath1 . despite this ,
the macroscopic , steady - state flow curve ( stress vs. strain rate ) is monotonic .
however , for a broad class of @xmath0 , the response to steady stress is not in fact steady flow , but spontaneous oscillation .
we discuss this finding in relation to other theoretical and experimental flow instabilities . within the parameter
ranges we studied , the model does not exhibit rheo - chaos .
the flow behaviour of shear - thickening materials such as dense colloidal suspensions can be complex @xcite .
for example , imposition of a steady mean strain rate can lead to large , possibly chaotic , variations in the mean stress @xcite . the same occurs in some types of shear - thickening micellar surfactant solutions , where true temporal chaos seems now to be established @xcite ( and also in shear thinning systems ; see @xcite ) .
other unexpected behaviour , such as a bifurcation to an oscillatory state , has also been seen in shear - thickening ` onion ' phases of surfactant @xcite .
it is not yet known to what extent such unsteady flow is generic in shear - thickening systems ; in this letter we attempt to shed some light on the issue by studying a much - simplified , generic model . in this model
we find , for a wide range of parameters , spontaneous rheological oscillation of the strain rate at fixed stress .
rheo - chaos is , however , not found for the parameters studied so far .
a feature that distinguishes the rheological instabilities encountered in shear - thickening from those arising in newtonian fluids is that the nonlinearity is not inertial ( not from the advective term of the navier stokes equation ) : the reynolds number is essentially zero @xcite .
instead it arises from anharmonic elastic responses at large deformations , complicated and perhaps strongly enhanced by the presence of memory effects .
flow instabilities leading to chaos have been studied recently by grosso _
et al . _
@xcite in a model for suspended rodlike particles . as that work shows , and our work confirms , temporal instabilities can arise even in a model where macroscopic spatial inhomogeneity is disallowed altogether .
this is a strong demarcation from the familiar shear - banding instabilities that arise whenever the steady - state flow curve is nonmonotonic ( the flow curve is the function @xmath2 , where @xmath3 is shear stress and @xmath4 the rate of shear strain ) @xcite .
shear banding is well documented in shear thinning materials ( particularly micelles @xcite ) but also possible in shear thickening ones @xcite , albeit with a different spatial organization of the bands of coexisting material .
however it is not the subject of this letter we are interested in _ temporal _ inhomogeneity . our simplified model of a shear thickening fluid is defined as follows .
we imagine an ensemble of mesoscopic elastic elements , each having a local strain variable @xmath1 .
we consider only simple shear strains , and neglect normal stresses , so that the only nontrivial stress is the corresponding shear stress @xmath3 .
( these assumptions reduce the problem to a scalar one . )
let us define @xmath5 to be the probability density function of elements which have a local strain @xmath1 at time @xmath6 .
we assume @xmath5 evolves in time according to two distinct mechanisms : homogeneous shearing at a rate @xmath7 , and the yielding of elements at a rate @xmath0 per unit time . in calculating the stress ,
the elements are supposed to behave elastically between yield events , so that the local stress is just @xmath8 , where @xmath9 is an elastic constant .
the global stress @xmath3 is simply the arithmetic mean of the local stresses , or @xmath10 .
( here the angled brackets represent an instantaneous average over @xmath5 . )
we ignore completely the fact that , in practice , the strain rate @xmath7 can vary locally in space in response to each mesoscopic element having a different local stress . in the present work ,
@xmath0 is the same for all elements , which simplifies the analysis significantly . a more comprehensive model , in which elements each have their own ` yield strain ' @xmath11 is studied in a longer paper @xcite , where further details of some of our calculations can also be found .
both models are closely related to the ( shear - thinning ) sgr model of sollich _ et al . _
@xcite , but we will choose a very different form of @xmath0 from that of sgr .
the master equation for @xmath5 is then @xmath12 the second term on the left hand side represents the increase in local strains @xmath1 according to the spatially uniform strain rate @xmath7 .
the two terms on the right hand side describe the birth and death of elements with strain @xmath1 , respectively , where @xmath13 reflects the assumption that newly
yielded elements are unstrained .
the total rate of yielding @xmath14 is defined by @xmath15 the key remaining ingredient of our model is the hopping rate @xmath0 .
we parameterise this as follows : @xmath16 \label{e : gamma}\ ] ] this is a pseudo - activated form , in which each element is subject to an effective temperature @xmath17 , and attempts ( at a rate @xmath18 ) to hop over a barrier of height @xmath19 .
the latter expression allows for the lowering of the elastic yield barrier by the imposition of strain ; on its own this will always lead to shear thinning .
however , the novel feature in the current model is to allow @xmath20 to be a decreasing function of @xmath1 .
this reflects the intuition that , even if a small local stress or strain always promotes yield , a large enough one may jam an element against its neighbours causing the jump rate to fall .
other than in steady state ( see below ) we have been unable to find an analytic solution to ( [ e : master ] ) for any non - trivial @xmath20 .
instead it has been numerically integrated . with the above as motivation , we chose for numerical purposes relatively simple ( piecewise constant ) decreasing functions @xmath20 .
the precise choice of @xmath20 in particular , whether it is smooth or not does not seem to qualitatively influence the results . |
a key ingredient in the understanding and modelling of galaxy evolution is the relationship between the large - scale star formation rate ( sfr ) and the physical conditions in the interstellar medium ( ism ) .
most current galaxy formation and evolution models treat star formation using simple ad hoc parametrizations , and our limited understanding of the actual form and nature of the sfr - ism interaction remains as one of the major limitations in these models ( e.g. , navarro & steinmetz 1997 ) .
measurements of the star formation law in nearby galaxies can address this problem in two important respects , by providing empirical recipes " that can be incorporated into analytical models and numerical simulations , and by providing clues to the physical mechanisms that underlie the observed correlations .
the most widely applied star formation law remains the simple gas density power law introduced by schmidt ( 1959 ) , which for external galaxies is usually expressed in terms of the observable surface densities of gas and star formation : @xmath2 the validity of the schmidt law has been tested in dozens of empirical studies , with most measured values of @xmath3 falling in the range 1 @xmath4 2 , depending on the tracers used and the linear scales considered ( kennicutt 1997 ) . on large scales the star formation law shows a more complex character , with a schmidt law at high gas densities , and a sharp decline in the sfr below a critical threshold density ( kennicutt 1989 , hereafter k89 ) .
these thresholds appear to be associated with large - scale gravitational stability thresholds for massive cloud formation ( e.g. , quirk 1972 ; fall & efstathiou 1980 ; k89 ) . at high gas densities , well above the stability threshold
, the form of the schmidt law appears to be remarkably consistent from galaxy to galaxy , both in terms of its slope ( @xmath5 ) and the absolute sfr efficiency ( the coefficient @xmath6 in eq . [ 1 ] ) .
studies of this kind offer the beginnings of a quantitative , physical prescription for the sfr that can be incorporated into galaxy formation and evolution models .
this is the first of two papers which reinvestigate the form and physical nature of the star formation law , over a much larger range of galaxy types and gas densities than was possible previously .
paper ii ( martin & kennicutt 1998 ) uses new ccd imaging of an hi and co selected sample of spiral galaxies to quantify the behavior of the star formation law within individual galaxies , and to test several models for the star formation law .
this paper is concerned with the behavior of the star formation law on global scales , averaged over the entire star forming disk .
such global laws , which treat galaxies in a single - zone approximation , provide less physical insight into the star formation process itself , but they provide very useful parametrizations ( recipes ) for galaxy evolution modelling .
earlier work has shown that the global , disk - averaged star formation law is reasonably well represented by a schmidt law ( k89 ; buat , deharveng , & donas 1989 ; buat 1992 ; boselli 1994 ; deharveng 1994 ) .
however these analyses have been hampered by small samples and by the small range of gas densities represented in those samples . in this paper
we use newly available hi , co , and data to more than double the sample over previous studies , and fully cover the range of mean gas densities found in disks .
we combine these data with published co , , and far - infrared ( fir ) measurements of luminous starburst galaxies , to investigate the nature of the schmidt law in higher density environments , thereby extending the total density range probed to nearly five orders of magnitude .
our main goal is to test whether the millionfold range in observed sfrs , extending from quiescent gas - poor disks to nuclear starbursts , can be understood within a common empirical and physical framework .
to investigate the global star formation law in normal disks , we searched the literature for normal galaxies with well - sampled hi and co measurements , and for which imaging or photometry are available .
our analysis of this sample closely follows that described in k89 . to investigate the star formation law at higher densities , we compiled published co maps , fir photometry , and emission - line measurements for a sample of infrared - selected starburst galaxies .
each data set is discussed separately below .
previous studies of the disk - averaged star formation law have shown that the global sfr correlates most strongly with the total ( atomic @xmath7 molecular ) gas density ( e.g. , kenney & young 1988 ; k89 ; buat 1992 ; boselli 1994 ) . consequently our primary data set consists of normal spirals for which spatially - resolved hi , co , and data are available .
a master list of candidate galaxies was compiled from the fcrao co survey ( young 1989 ; 1995 ) , supplemented by the co survey of sage ( 1993 ) . within these samples
, we identified 61 galaxies which also have published hi maps , photometry , and inclinations less than 75 ( to avoid severe extinction problems in edge - on systems ) .
total hi masses based on single - dish measurements are available for another 150 galaxies , but those data are unsuitable for the current application , because much of the hi is located well outside of the star forming disks , and it is essential to correlate the sfr and gas densities over the same physical region .
however we do use some of these additional galaxies in 3.1 to examine the form of the sfr _ vs _ hi schmidt law .
table 1 lists the 61 galaxies in the sample , the relevant surface densities , and references , as described below .
when considering the sample properties as a whole the main selection criterion was availability of co and hi maps , so the galaxies should comprise a virtually unbiased set in terms of star formation properties .
approximately 40% of the galaxies are members of the virgo cluster , selected from the co survey of kenney & young ( 1988 ) and the hi survey of warmels ( 1988 ) , and this sample contains most of the luminous spirals in the cluster core .
the field galaxy subsample is more heterogeneous , and is significantly biased toward galaxies of hubble type sb and later , but it is unlikely that this selection biases the form of the star formation law .
hi surface densities were taken mainly from the compilations of warmels ( 1988 ) , broeils & van woerden ( 1994 ) , and rhee & van albada ( 1995 ) , supplemented by individual measurements of a few galaxies ( table 1 ) .
the mean hi surface densities , averaged within the optical radius of the disk , were derived from the surface density profiles given in those papers or the references therein .
the disk radii are the corrected isophotal radii as given in the rc2 catalog ( de vaucouleurs , de vaucouleurs , & corwin 1976 ) . the mean densities used here differ from those that are often tabulated in the original papers , the latter usually being averaged within the inner _ half _ of the optical disk .
total molecular hydrogen masses were taken from the young ( 1989 ; 1995 ) and sage ( 1993 ) surveys , and converted when needed to a common co / h@xmath1 conversion factor : @xmath8 @xmath9 ( k km s@xmath10 ) .
the mean 2 surface densities were then determined , by averaging within the radii listed in table 1 .
these average densities are meaningful only if the co emission is confined to the optical disk , and the measurements extend to a substantial fraction of optical radius .
galaxies which were sampled to less than half of the optical radius were not included in our sample .
integrated sfrs were derived from measurements of the emission - line flux , following the method described in kennicutt ( 1983 ) .
most of the fluxes were taken from the surveys of kennicutt & kent ( 1983 ) , romanishin ( 1990 ) , and young ( 1996 ) .
those data were supplemented with new calibrated ccd images obtained with a focal reducer camera on the steward observatory 2.3 m bok telescope , and with the 0.9 m and burrell schmidt telescopes at kitt peak national observatory .
details of these observations are given in paper ii .
the fluxes were corrected as needed for foreground extinction and [ nii ] emission , following the prescriptions in kennicutt ( 1983 ) .
the original fluxes of kennicutt & kent ( 1983 ) have been corrected upwards by a factor of 1.16 to place them on a consistent zeropoint with more recent measurements ( romanishin 1990 ; kennicutt 1992 ) .
the luminosities were then converted to total sfrs , using the updated calibration of kennicutt , tamblyn , & congdon ( 1994 ) : @xmath11 the luminosities used in equation ( 2 ) were corrected for internal extinction by 1.1 mag ( factor 2.8 ) , based on a comparison of free - free radio fluxes and fluxes of galaxies by kennicutt ( 1983 ) .
the actual extinction varies within the sample , of course , which introduces significant scatter in the observed star formation law , as discussed later .
while it would be much better to apply individual extinction corrections to each galaxy , determining the reddening or extinction from integrated spectra is problematic ( kennicutt 1992 ) , and would introduce uncertainties that are larger than the single average correction .
it may be possible in the future to derive improved estimates of the extinction and sfr using measurements of near - infrared brackett or paschen recombination lines , but such data are not currently available .
the imf used in this conversion is a salpeter function ( @xmath12 ) over @xmath13 .
the salpeter imf was adopted in order to be consistent with the infrared - derived sfrs in the next section .
adopting the extended miller - scalo function used in kennicutt ( 1983 ) would produce nearly identical sfrs ( only 8% lower ) .
galaxy distances from young ( 1989 ) were used in this intermediate calculation , but the distances are irrelevant for most of this paper , because the schmidt law is analyzed in terms of distance - independent surface densities . finally , the mean sfr surface density ( units yr@xmath10 kpc@xmath14 ) was derived for each galaxy , by dividing the total sfr from equation ( 2 ) by the deprojected area within the corrected rc2 radius . through the remainder of this paper , we shall refer to this sfr per unit area as the sfr density " .
in most galaxies the rc2 radius coincides approximately with the edge of the main -emitting disk ( k89 ) , so the sfr density as measured here corresponds roughly to the mean sfr per unit area within the active star forming disk .
the derived sfr surface densities are listed in table 1 . the observed surface densities ( uncorrected for extinction ) can be derived from table 1 by the simple relation : @xmath15 , where @xmath16 is expressed in units of ergs sec@xmath10 pc@xmath14 .
this conversion may be useful for readers who may wish to apply a different sfr calibration to the data compiled here .
the mean gas densities of the normal spiral disks in our sample lie within a relatively narrow range , from 2 to 50 pc@xmath14 , and this seriously limits the dynamic range over which the behavior of the schmidt law can be evaluated .
the density range can be extended to @xmath17100 pc@xmath14 by analyzing spatially - resolved measurements of individual disks ( paper ii ) , but above these densities measurements become unreliable for determining the sfr . for a typical gas - to - dust ratio found in nearby galaxies , the visual extinction reaches 1 mag for column densities @xmath18 @xmath9 , or @xmath19
pc@xmath14 ( e.g. , bohlin , savage , & drake 1978 ; caplan & deharveng 1986 ) .
hence one expects the extinction at to become problematic for regions with mean gas surface densities above 50
100 pc@xmath14 .
if we wish to study the nature of the star formation law in these dense regions , a star formation diagnostic other than must be used .
large - scale star formation at much higher densities is commonly found in the centers of normal galaxies , and particularly in luminous infrared starburst galaxies . in order to analyze the star formation law in this regime
, we searched the literature for high - resolution co and infrared measurements of starburst galaxies . since the starbursts are often concentrated in compact circumnuclear disks ( e.g. , scoville 1994 ; sanders & mirabel 1996 ; smith & harvey 1996 ) , high - resolution data are required in order to accurately determine the linear sizes of the starburst regions and the corresponding surface densities .
our sample comprises 36 galaxies with high - resolution co data , most based on aperture synthesis mapping , and for which infrared measurements of the same region are available .
the sample ranges from low - level nuclear starbursts in normal and peculiar galaxies such as ngc 253 , ic 342 , maffei 2 , and m82 ( @xmath20 ) to ultraluminous starburst galaxies with @xmath21 ( e.g. , arp 220 ) .
care was taken to select objects in which the dust heating is dominated by a starburst , as determined from optical spectra spectra ( e.g. , armus , heckman , & miley 1989 ; veilleux 1995 ) and/or mid - infrared spectroscopy ( e.g. , lutz 1996 ) . objects with evidence for a strong agn component were excluded ( e.g. , ngc 1068 , ngc 7469 , mrk 231 , mrk 273 ) .
total molecular gas masses in the starburst disks were derived from the co flux and distance , using the same co/2 conversion factor as for the normal galaxies .
the validity of a constant conversion factor is highly questionable ( e.g. , wild 1992 ; downes , solomon , & radford 1993 ; aalto 1994 ; solomon 1997 ) , and we have adopted a uniform conversion factor strictly for the sake of simplicity .
the impact of adopting a different conversion factor will be discussed later .
the mean molecular surface densities were then derived , averaged within the radius of the central molecular disk as determined from the co maps .
high - resolution hi observations are only available for a few of these galaxies , and in those cases the atomic fraction in the circumnuclear region is small , of order a few percent or less ( e.g. , garcia - barreto 1991 ; downes 1996 ; sanders & mirabel 1996 ) .
this is not surprising given the very high column densities found in these regions .
consequently we have ignored the hi component and approximate the molecular mass as the total gas mass in the starburst region .
table 2 lists the galaxies in the sample , the radii of the disks , and their mean molecular surface densities .
the sfrs for the starbursts were derived from measurements of their fir luminosities .
these were taken from a variety of sources , as listed in table 2 .
for about half of the sample , high - resolution maps at mid - infrared wavelengths are available , and when combined with iras fluxes for the galaxies as a whole they provide an accurate estimate of the fir luminosity in the central starbursts themselves ( telesco , dressel , & wolstencroft 1993 ; smith & harvey 1996 ) . for the other galaxies the fir luminosity of the starburst
was derived from a combination of iras photometry and groundbased aperture photometry at 1020 , or from the iras fluxes alone , in cases where most of the total fir emission appears to originate in the central starburst .
sfrs for three of the galaxies were derived from a combination of and infrared photometry , as noted in table 2 . in normal disk galaxies
the relationship between the fir luminosity and the sfr is complex , because stars with a variety of ages can contribute to the dust heating , and only a fraction of the bolometric luminosity of the young stellar population is absorbed by dust ( e.g. , lonsdale & helou 1987 ; walterbos & greenawalt 1996 ) .
however in the starbursts studied here , the physical coupling between the sfr and the ir luminosity is much more direct .
young stars dominate the radiation field that heats the dust , and the dust optical depths are so large that almost all of the bolometric luminosity of the starburst is reradiated in the infrared .
this makes it possible to derive a reasonable quantitative measure of the sfr from the fir luminosity .
our calibration of the sfr/@xmath22 conversion is based on the starburst synthesis models of leitherer & heckman ( 1995 ) .
their models trace the temporal evolution of the bolometric luminosity for a fixed sfr , metal abundance , and imf .
we computed the sfr calibration using their continuous star formation " models , in which the sfr is presumed to remain constant over the lifetime of the burst .
the models show that the l@xmath23/sfr ratio evolves relatively slowly between ages of 10 and 100 myr , the relevant range for most of these starbursts ( e.g. , bernlhr 1993 ; engelbracht 1997 ) .
alternatively one can derive the conversion using a instantaneous burst " approximation , where it is assumed that star formation has ceased , but the calibration is sensitive to the presumed burst age and the ( questionable ) assumption of an instantaneous burst . adopting the mean luminosity for 10100 myr continuous bursts , solar abundances , the salpeter imf described earlier , and assuming that the dust reradiates all of the bolometric luminosity yields : @xmath24 this lies within the range of previously published calibrations ( @xmath25 yr@xmath10 @xmath26 .
equation ( 3 ) yields sfrs that are 14% lower than the recent calibration of lehnert & heckman ( 1996 ) , and 22% lower than meurer ( 1997 ) .
the sfr surface density was then calculated within the radius of the starburst region as determined from the co maps , or from the infrared maps if high - resolution co data were not available .
the sizes of the regions defined in co and the infrared show excellent correspondence in cases where comparable resolution data are available ( telesco 1993 ; smith & harvey 1996 ) .
table 2 lists the radii , gas densities , and sfr surface densities derived in this way . in 4
we analyze the composite properties of the normal disk and starburst samples , so it is important to confirm that the fir and -based sfrs are on a consistent zeropoint .
matching aperture photometry for 18 of the galaxies in our sample is available from the compilations of puxley , hawarden , & mountain ( 1990 ) , telesco ( 1993 ) and smith & harvey ( 1996 ) , and these allow us to compare the emission - line and fir sfr scales on a self consistent basis .
the fir - based sfrs were derived using equation ( 3 ) , while the -based sfrs were derived using equation ( 2 ) and a / ratio of 0.0103 , corresponding to case b recombination at @xmath27 k and @xmath28 @xmath29 ( osterbrock 1989 ) .
no extinction corrections were applied to the data .
figure 1 shows a comparison of the fir and -derived sfrs .
the solid line shows the correlation expected if the two sets of sfrs were equivalent .
the data in figure 1 closely follow this correlation , but the fir - derived sfrs are systematically higher by an average of [email protected] dex , as shown by the dashed line .
this displacement could indicate a general inconsistency between the zeropoints of the and fir calibrations of the sfr , which might arise , for example , from errors in the fir luminosities ( many of them extrapolated from the mid - ir ) , or in the synthesis model that is used to convert the fir luminosities to sfrs . however there is physical justification for expecting that the fluxes would systematically underestimate the sfrs in many of these objects .
the extinction in most regions is so large that one expects part of the ionizing radiation from the starburst to be absorbed by grains , and in some objects extinction of itself is probably significant ( e.g. , lutz 1996 ; goldader 1997 ) .
the -derived sfr will also tend to be systematically lower than the fir - derived value if the starbursts are observed after the peak of the burst , because the dust heating is dominated by longer lived stars than the emission lines .
we provisionally adopt the sfrs from equation ( 3 ) in the following analysis , on the tentative assumption that the fir - based sfrs are more reliable in these objects .
however we will also explore the consequences of adopting the lower -based scale , and include this uncertainty in the analysis of the global schmidt law .
individual uncertainties are not listed for the surface densities listed in tables 1 and 2 , because the predominant errors are systematic in nature and difficult to quantify .
however it is important to be aware of nature of these uncertainties and their possible influence on the observed star formation law . for the normal spiral disks , with sfrs derived from luminosities ( table 1 ) , the dominant systematic errors are extinction variations , which introduce a scatter in the sfr densities , and uncertainty in the extrapolated imf , which could introduce an overall shift in the sfrs ( kennicutt 1983 ) .
the dominant errors in the gas densities are variation in the co / h@xmath1 conversion factor , combined with the limited sampling of the co measurements in some galaxies ( sage 1993 ; young 1995 ) .
a realistic estimate for the observational scatter in the sfrs is @xmath303050% , or @xmath300.150.3 dex ( kennicutt 1983 ) , and the uncertainties in the gas densities are probably comparable . we adopt an average uncertainty of @xmath300.2 dex in the following analysis .
the systematic uncertainties in the sfrs and gas densities derived for the starburst galaxies ( table 2 ) are larger . in many cases
the fir luminosities have been derived from a combination of high - resolution mid - infrared measurements and iras fir fluxes , and there can be substantial uncertainty in the extrapolation to a total fir flux .
in other cases only integrated iras fluxes for the galaxies are available , and the presence of significant fir emission from the region outside of the central starburst will cause the starburst sfr to be systematically overestimated .
the sfr will also be overestimated if the dust is heated partly by other sources , such as an active nucleus .
another significant source of uncertainty in the sfrs inferred for individual starbursts is the use of a fixed continuous burst model , though the effect on the overall sfr scale should be lower .
the gas densities in the starburst regions are also subject to systematic error as well , mainly through uncertainties in the co / h@xmath1 conversion factor ( e.g. , downes 1993 ; solomon 1997 ) .
other smaller sources of uncertainty include the neglect of atomic gas and errors in the radii of the starbursts .
the latter errors affect the inferred sfr and gas densities equally , and have less of an effect on the form of the schmidt law .
the largest of these systematic uncertainties , the l@xmath31 sfr conversion and the co/2 conversion , could introduce errors in the sfr or gas density scales at the factor of 2 @xmath4 3 level ( 0.3 @xmath4 0.5 dex ) . in our analysis
we adopt uncertainties @xmath32 dex in both parameters , with the asymmetry reflecting the greater likelihood that the systematic errors tend to lead to overestimates of the sfrs and gas densities . despite these uncertainties ,
the data provide very strong constraints on the form of the star formation law , because of the very large range of absolute densities and sfrs represented in the sample , 2 @xmath4 6 orders of magnitude depending on the subsample of interest .
figure 2 shows the relationship between the disk - averaged sfr and total gas density ( atomic and molecular hydrogen ) for the 61 normal spirals in our sample . a clear correlation is apparent in the expected sense of increasing sfr with increasing gas densities , with a mean slope that is considerably steeper than a linear relation ( indicated by the dotted and dashed lines ) .
however the scatter in the relation is large , up to a factor of 30 in sfr at a fixed gas density , and comparable to the total range in observed gas density .
consequently the slope of the schmidt law is poorly constrained . a conventional least squares fit which minimizes ( logarithmic ) residuals in the sfr density yields @xmath33 .
this slope lies in the middle of the range @xmath34 derived in previous studies with smaller samples ( buat 1989 ; k89 ; buat 1992 ; deharveng 1994 ) .
a bivariate least squares regression , which takes into account the uncertainties in the gas densities as well , yields a much steeper fit @xmath35 .
both fits are shown with solid lines in figure 2 .
the large difference between these solutions is a direct reflection of the large dispersion in the disk - averaged sfr vs gas density relation , and the result underscores the conclusion that any schmidt law in these galaxies should be regarded as a _ very _ approximate parametrization at best .
what is the physical origin of the large dispersion in figure 2 ? as discussed earlier , variations in extinction and the co/2 conversion introduce a scatter at about the @xmath300.2 dex level in the sfr and gas densities , as signified by the error bars in figure 2 .
this can account for roughly half of the observed scatter in the star formation law .
the remaining scatter must be real , reflecting a real variation in the mean schmidt law .
such a variation is not entirely surprising , when one recalls that the local sfrs and gas densities span orders of magnitude within typical disks , and averaging over the entire disk will not necessarily preserve the form of a nonlinear local schmidt law .
the problem is illustrated in figure 3 , which shows the radial sfr vs gas density profiles for 21 of the galaxies in our sample ( paper ii ) .
each profile was produced by measuring the azimuthally averaged gas density and sfr density as a function of galactocentric radius , then plotting the resulting sfr vs gas density relation on a common scale . at high densities the sfrs
are well represented by a shallow schmidt law ( @xmath36 ) , but the slope of the star formation law steepens abruptly below the threshold density .
the disk - averaged sfrs plotted in figure 2 represent gross averages over these highly nonlinear relations , and the resulting global schmidt law exhibits a slope that is intermediate between the @xmath36 power - law dependence at high density and the steeper law in the threshold regime . the dispersion in figure 2 is introduced because the star formation in some galaxies is highly concentrated to the high - density part of the local schmidt law , while in other systems much of the star formation takes place near the threshold density ( see k89 ) .
this underscores the caveat that disk - averaged schmidt law analyzed here contains little physical information about the underlying star formation law .
however it does provide a convenient means of parametrizing the gross star formation properties of disks in simple one - zone evolution models .
we defer further discussion of the spatially - resolved star formation law for paper ii .
the data in figure 2 also provide useful information on the average global efficiency of star formation in local disks , the coefficient @xmath6 in equation ( 1 ) .
the dashed and dotted lines in figure 2 correspond to constant sfrs per unit gas mass , in units of 1% , 10% , and 100% per @xmath37 yr .
the choice of @xmath37 yr as a fiducial timescale is arbitrary , but it does correspond roughly to a typical orbital time in the disks .
the lines are offset by a factor of 1.37 to include helium and heavy elements in the total gas mass .
the median efficiency for the disks in figure 2 is 4.8% , i.e. , a typical present - day spiral galaxy converts 4.8% of the gas ( within the optical radius ) to stars over this period .
the efficiencies can be expressed alternatively as gas consumption timescales , with the lines in figure 2 corresponding to timescales @xmath38 of 10 , 1 , and 0.1 gyr ( bottom to top ) .
the median gas consumption time for the disks in this sample is 2.1 gyr , again referring to the star forming disks alone , and not including corrections for recycling of interstellar gas .
recycling typically extends the actual consumption timescale by factors of 23 above the simple calculation ( kennicutt 1994 ) .
most of the galaxies in figure 2 possess disk - averaged star formation efficiencies in the range 2 @xmath4 10% per @xmath37 yr , corresponding to gas consumption times of 1 @xmath4 5 gyr .
however several galaxies are more extreme , and the full range of efficiencies is 0.8 @xmath4 60% per @xmath37 yr ( @xmath38 = 0.2 @xmath4 12 gyr ) .
the shortest timescales correspond to optically - selected starburst galaxies such as ngc 1569 and ngc 3310 , while the low extremes are represented by early - type spirals such as m31 , ngc 2841 , and ngc 4698 , where the current sfrs are so low that the future consumption times , even for their modest gas supplies , are comparable to the hubble time . until now our attention has focussed solely on the relationship between the disk - averaged sfr and the total gas density , but we can also examine how the sfrs correlate with the average atomic and molecular gas densities , as shown in figure 4 .
these comparisons include galaxies mapped in hi or co ( but not both ) , so the samples are considerably larger than shown in figure 2 .
the left panel of figure 4 shows the sfr vs hi density relation for 88 galaxies with and hi data in common .
the correlation is very reminiscent of the sfr vs total density relation shown in figure 2 , and in fact the correlation coefficients are nearly identical , 0.66 for the sfr @xmath4 hi relation vs 0.68 for the sfr @xmath4
hi@xmath72 relation .
this is not entirely surprising , as hi accounts for approximately half of the total gas density on average .
these results are consistent with previous analyses based on smaller samples by k89 , buat ( 1992 ) , deharveng ( 1994 ) , boselli ( 1994 ) , and boselli ( 1995 ) .
the physical interpretation of the sfr vs hi schmidt law is not obvious , however .
it may trace the physical influence of the atomic gas density on the sfr , but it could be that the sfr regulates the density of hi , through the photodissociation of molecular gas by hot stars ( shaya & federman 1987 ; tilanus & allen 1989 ) .
the correlation between the -based sfrs and h@xmath1 density is much weaker , as shown in the right panel of figure 4 .
this has been reported previously , and appears to hold independently of whether sfrs based on , uv continuum fluxes , or fir fluxes are analyzed ( buat 1992 ; boselli 1994 ) .
such a poor correlation between the sfr and molecular gas densities is unexpected , and it has led some to suggest that variations in the co / h@xmath1 conversion factor are responsible for the scatter ( k89 ; boselli 1994 ; boselli 1995 ) .
our data provide indirect support for this interpretation .
several lines of evidence suggest that the galactic co/2 conversion factor is valid in regions with near - solar metallicity , but that it tends to systematically underestimate the h@xmath1 mass in metal - poor regions , such as are found in the outer disks of spirals or in low - luminosity galaxies ( e.g. , maloney & black 1988 ; kenney & young 1988 ; rubio 1993 ; wilson 1995 ) . to test
whether this effect might be contributing to the scatter in figure 4 , we subdivided our sample by blue luminosity , with solid points denoting galaxies with l@xmath39 ( @xmath40 for h@xmath41 = 75 km s@xmath10 mpc@xmath10 ) and open circles representing fainter galaxies .
the mean metal abundance in disks is well correlated with luminosity , so this provides an approximate separation of the galaxies by abundance , around a value of @xmath42 ( zaritsky , kennicutt , & huchra 1994 ) .
figure 4 shows that the luminous , metal - rich spirals do show a much better defined sfr vs h@xmath1 density correlation , comparable in slope and scatter to the correlations with total and hi density .
by contrast , the low - luminosity galaxies show essentially no correlation between the sfr and co - inferred h@xmath1 densities , with many co - weak galaxies showing unusually _ high _ sfrs .
although this is hardly a conclusive result , it offers circumstantial evidence that variations in the co / h@xmath1 conversion factor are responsible for most of the scatter in the sfr vs molecular gas density relation .
our conclusions are consistent with those of boselli ( 1994 ) and boselli ( 1995 ) , and the reader is referred to those papers for more detailed discussions of this problem .
we can perform a parallel analysis for the infrared - selected starbursts , and the results are summarized in figure 5 .
the comparison is directly analagous to that shown for the normal disks in figure 2 , except that the sfrs are derived from fir luminosities , and the sfrs are correlated with the h@xmath1 gas density alone ( the disks are expected to be overwhelmingly molecular , as discussed earlier ) .
the sfrs and densities are averaged within the radii of the central molecular disks and starbursts , which have typical dimensions of order 1 kpc .
the error bars indicate the typical uncertainties , as discussed in 2.3 .
the starburst galaxies also show a well - defined schmidt law , in this case with a best fitting least squares slope @xmath43 ( bivariate regression ) or @xmath44 ( errors in sfrs only ) .
the schmidt law is better defined than for the normal disks , but partly because there is a much larger dynamic range in sfr and gas densities in the starburst sample ; the dispersion in absolute sfr per per unit area at fixed gas density is only slightly lower in the starburst sample .
star formation threshold effects are probably unimportant in the starburst disks , and this might also account for the somewhat tighter schmidt law among these objects . although the starburst disks exhibit a sfr vs gas density relation that is qualitatively similar in form to that seen in the normal spiral disks , the physical regime we are probing is radically different .
the average gas surface densities here range from @xmath45 to @xmath46 pc@xmath14 , compared to a typical range of order 1 @xmath4 100 pc@xmath14 in normal disks ( figures 2 , 3 ) .
the mean densities of the starburst disks are comparable instead to those of individual molecular cloud complexes in normal galaxies .
for example ,
the largest hii / gmc complexes in m31 , m33 , and m51 have molecular masses and sizes corresponding to mean surface densities of 40 @xmath4 500 pc@xmath14 ( wilson & rudolph 1993 ; wilson & scoville 1992 ; nakai & kuno 1995 ) .
this is comparable to the _ low end _ of the density range for the starbursts in figure 5 .
the mean densities of some of the starbursts approach those of galactic molecular cloud cores , but extending over kiloparsec diameter regions .
the star formation densities are just as extraordinary .
for example , the central 10 pc core of the 30 doradus giant hii region contains @xmath47 in young stars , which corresponds to @xmath48 yr@xmath10 kpc@xmath14 if the star formation timescale is as short as @xmath49 yr ; the average sfr density averaged over the entire hii region is @xmath171 @xmath4 10 yr@xmath10 kpc@xmath14 .
thus the regions we are studying have projected sfrs per unit area that approach the maximum limit observed in nearby optically - selected star clusters and associations ( meurer 1997 ) , but extending over regions up to a kiloparsec in radius . not surprisingly , the global star formation efficiencies in the starburst sample are much higher than in the normal disk sample ( e.g. , young 1986 ; solomon & sage 1988 ; sanders , scoville , & soifer 1991 ) . in figure 5 we show the same lines of constant star formation efficiency and gas consumption times as in figure 2 ( 1% , 10% , and 100% per @xmath37 yr ) .
the median rate of gas consumption is 30% per 10@xmath50 yr , 6 times larger than for the normal disk samples , and the efficiencies reach 100% per @xmath37 yr for the most extreme objects .
it is interesting to note that the shortest gas consumption times are comparable to the dynamical timescales of the parent galaxies , implying that the most luminous starbursts are forming stars near the limit set by the gas accumulation timescale ( lehnert & heckman 1996 ) .
taken together , the normal disk and starburst samples span a dynamic range of approximately @xmath46 in gas surface density and over @xmath49 in sfr per unit area .
figure 6 shows the composite relation , with the normal spirals shown as solid circles and the starbursts as solid squares .
quite remarkably , the data are consistent with a common schmidt law extending over the entire density range .
figure 6 shows that the normal disk and starburst samples occupy completely separate regimes in gas density and sfr per unit area , not a surprising result given the very different selection criteria for the two samples .
but before we interpret the composite relation it is important to establish whether there is a smooth physical continuity between the normal disk and starburst regimes , and to confirm the consistency of the and fir - derived sfr scales .
to this end we derived -based sfrs and gas densities for the central regions of 25 of the normal spirals in table 1 ( @xmath51 ) , using our images and published hi and co maps ( paper ii ) .
the resulting sfr and gas densities are shown as open circles in figure 6 .
these regions span the physical parameter space between the normal disks as a whole and the infrared - selected circumnuclear starburst regions .
figure 6 shows that the gas densities and -derived sfrs of these regions fall on the composite schmidt law defined by the normal disk and starburst samples , and fill the transition region between the two physical regimes .
the same conclusion can be drawn by comparing the sfrs of the infrared - selected starburst galaxies in figure 5 with the spatially - resolved sfrs of the normal disks shown in figure 3 ; the starbursts lie on the extrapolation of the high - density star formation laws observed in the spiral disks .
this result , combined with the -fir comparison discussed earlier , gives us confidence that we are measuring the form of the star formation law on a self - consistent basis across the sample .
the solid line in figure 6 shows a bivariate least - square fit to the composite relation defined by the normal disks and the starbursts ( but not including the open circles ) . in this case
we applied equal weights to all of the data points , in order to avoid having the fit driven by the normal spirals in the lower left region of figure 6 .
this yields a best fitting index @xmath52 ( bivariate regression ) or @xmath53 ( errors in sfrs only ) . these are nearly identical to the schmidt law fits for the starburst sample alone , which further confirms the consistency of the large - scale star formation laws in the two samples .
the formal uncertainties listed here assume random errors of @xmath300.3 dex in the sfrs and gas densities , but it underestimates the full uncertainty in the schmidt law , because we have not accounted for the possibility of a systematic shift in the overall sfr or density scales for the starburst sample as a whole .
the effect of such a shift is easily calculated .
for example , reducing the sfrs for all of the starbursts by a factor of two , to match the calibration in figure 1 , would lower the best fitting index @xmath3 from 1.40 to 1.28 .
likewise , lowering the gas masses in the starbursts by a factor of two , to take into account the possibility that the co/2 conversion factor is systematically lower , would _ increase _
@xmath3 by approximately the same amount , from 1.4 to 1.5 .
this range of values provides a fairer estimate of the actual uncertainty in the composite schmidt law .
folding together all of these uncertainties , we adopt as our final result : @xmath54 figure 6 shows that equation ( 4 ) provides an excellent parametrization of the global sfr , over a density range extending from the most gas - poor spiral disks to the cores of the most luminous starburst galaxies .
this may account for why conventional galaxy evolution models , which usually are based on a schmidt law parametrization of the sfr , often produce realistic predictions of the gross star formation properties of galaxies .
there are limitations to the schmidt law in equation ( 4 ) that should be borne in mind , however , when applying this recipe to galaxy evolution models or numerical simulations .
although the full range of sfrs and gas densities are very well represented by a single power law with @xmath55 , the scatter in sfrs about the mean relation is substantial , @xmath300.3 dex rms , and individual galaxies deviate by as much as a factor of 7 .
consequently equation ( 4 ) provides at most a statistical description of the global sfr , averaged over large samples of galaxies .
another potential limitation for its application to simulations and models is the need to accurately specify the linear sizes of the relevant star forming regions .
this is relatively straightforward for normal disks , where the scaling radius is comparable to the photometric radius of the galaxy or the edge of the active star forming disk
. it may be more difficult to model in starbursts , however , where the intense star formation is usually concentrated in a region that is a few percent of the radius of the parent galaxy .
fortunately the slope of the schmidt law is relatively shallow , and a modest error in the scaling radius will displace the inferred sfr and gas densities nearly along a line of slope @xmath56 , nearly parallel to the schmidt law itself .
this is illustrated in figure 6 , where a short diagonal line shows the effect of changing the scaling radius by a factor of two ( for a fixed gas mass and total sfr ) .
the schmidt law in figure 6 is so well defined that it is tempting to identify a simple , unique physical origin for the relation .
however we find that a schmidt law is not the only simple parametrization that can reproduce the range of sfrs observed in this sample , and this serves as a caution against overinterpreting the physical nature of the empirical star formation law . in this section
we briefly discuss the form of the schmidt law expected from simple gravitational arguments , and demonstrate that a simple kinematical model provides an equally useful recipe for modelling the large - scale sfr .
numerous theoretical scenarios which produce a schmidt law with @xmath3 = 1 @xmath4 2 can be found in the literature ( larson 1992 and references therein ) .
simple self - gravitational models for disks can reproduce the large - scale star formation thresholds observed in galaxies ( quirk 1972 ; k89 ) , and the same basic model is consistent with a schmidt law at high densities with index @xmath57 ( larson 1988 , 1992 ) .
for example in a simple self - gravitational picture in which the large - scale sfr is presumed to scale with the growth rate of perturbations in the gas disk , the sfr will scale as the gas density divided by the growth timescale : @xmath58 where @xmath59 and @xmath60 are the volume densities of gas and star formation .
the corresponding scaling of the projected surface densities will depend on the scale height distribution of the gas , with @xmath61 expected for a constant mean scale height , a reasonable approximation for the galaxies and starbursts considered here .
although this is hardly a robust derivation , it does show that a global schmidt law with @xmath57 is physically plausible . in a variant of this argument , silk ( 1997 )
has suggested a generic form of the star formation law , in which the sfr surface density scales with the ratio of the gas density to the local dynamical timescale : @xmath62 where @xmath63 refers in this case to the local orbital timescale of the disk , and @xmath64 is the angular rotation speed .
models of this general class have been studied previously by wyse ( 1986 ) and wyse & silk ( 1989 ) , though with different scalings of the gas density and separate treatment of the atomic and molecular gas .
equation ( 6 ) might be expected to hold if , for example , star formation triggering by spiral arms or bars were important , in which case the sfr would scale with orbital frequency .
to test this idea , we compiled rotation velocities for the galaxies in tables 1 and 2 , and used them to derive a characteristic value of @xmath63 for each disk .
the timescale @xmath63 was defined arbitrarily as @xmath65 , the orbit time at the outer radius @xmath66 of the star forming region .
the mean orbit time in the star forming disk is smaller than @xmath63 defined in this way , by a factor of 1 @xmath4 2 , depending on the form of the rotation curve and the radial distribution of gas in the disk .
we chose to define @xmath63 and @xmath64 at the outer edge of the disk to avoid these complications .
tables 1 and 2 list the adopted values , in units of @xmath37 yr .
face - on galaxies or those with poorly determined ( rotational ) velocity fields were excluded from the analysis .
figure 7 shows the relationship between the observed sfr density and @xmath67 for our sample .
the solid line is not a fit but simply a line of slope unity which bisects the relation for normal disks .
this alternate prescription for the star formation law provides a surprisingly good fit to the data , both in terms of the slope and the relatively small scatter about the mean relation .
when compared over the entire density range the observed law is slightly shallower than predicted by equation ( 7 ) ( slope @xmath170.9 instead of 1 ) ; on the other hand the fit to the normal disk sample is as tight as a schmidt law .
the zeropoint of the line corresponds to a sfr of 21% of the gas mass per orbit at the outer edge of the disk .
since the average orbit time within the star forming disk is about half that at the disk edge , this implies a simple parametrization of the local star formation law : @xmath68 from a strictly empirical point of view , the schmidt law in equation ( 4 ) and the kinematical law in equation ( 7 ) offer two equally valid parametrizations for the global sfrs in galaxies , and either can be employed as a recipe in models and numerical simulations .
it is unclear whether the kinematic model can fit the radial distribution of star formation as well as a schmidt law , and we plan to explore this in paper ii .
the two parametrizations also offer two distinct interpretations of the observation that the star formation efficiency in central starbursts is much higher than found in quiescent star forming disks ( e.g. , young 1986 ; solomon & sage 1988 ; sanders 1991 ) . in the schmidt law picture ,
the higher efficiencies in starbursts are simply a consequence of their much higher gas densities . for a given index @xmath3
, the sfr per unit gas mass will scale as @xmath69 , and hence for the law observed here roughly as @xmath70 .
the central starbursts have characteristic gas densities that are 100 @xmath4 10000 times higher than the average for normal disks , hence we would expect the global star formation efficiencies to be 6 @xmath4 40 times higher , as observed . in the alternative picture in which the sfr is presumed to scale with @xmath67 , the high sfrs and star formation efficiencies in starburst
galaxies simply reflect the smaller physical scales and shorter dynamical timescales in these compact central regions .
it is difficult to to differentiate between these alternatives with disk - averaged measurements alone , and since the global star formation law is mainly useful as an empirical parametrization , the distinction may not be important .
deeper insight into the physical nature of the star formation law requires spatially resolved data for individual disks , of the kind that will be analyzed in paper ii .
several individuals contributed to the large set of data analyzed in this paper , and it is a pleasure to thank them .
the kpno data used in this paper were obtained as part of other projects in collaboration with r. braun , r. walterbos , and p. hodge .
c. martin worked on the reduction of the spatially resolved data shown in figure 3 .
i am also grateful to j. black , j. ostriker , s. sakai , p. solomon , s. white , and especially j. silk for comments and suggestions about early versions of this work .
i am also grateful to the anonymous referee for several comments that improved the paper .
some of the data used in this paper were obtained on the 2.3 m bok telescope at steward observatory .
this research was supported by the national science foundation through grant ast-9421145 .
fig . 1. a comparison of integrated sfrs derived from emission - line fluxes and far - infrared continuum luminosities , for 18 infrared - selected starburst galaxies .
the solid line shows the relation expected from eqs .
( 2 ) and ( 3 ) .
the dashed line is the best fitting mean relation .
2. relation between the disk - averaged sfr per unit area and gas density for 61 normal disk galaxies .
the solid lines are least square fits to the schmidt law , as described in the text .
the dashed and dotted lines correspond to constant global star formation efficiencies and gas consumption timescales , as indicated . fig .
4. correlation of the disk - averaged sfr per unit area with the average surface densities of hi ( left ) and 2 ( right ) .
the 2 densities were derived using a constant co/2 conversion factor . in the right panel ,
solid circles denote galaxies with l@xmath39 , while open circles denote galaxies with l@xmath71 .
5. relation between the disk - averaged sfr per unit area and molecular gas density for 36 infrared - selected circumnuclear starbursts .
the solid line shows a bivariate least squares fit to the schmidt law , as described in the text .
the dashed and dotted lines correspond to constant global star formation efficiences and gas consumption timescales , as indicated . fig .
6. composite star formation law for the normal disk ( solid circles ) and starburst ( squares ) samples .
open circles show the sfrs and gas densities for the centers of the normal disk galaxies .
the line is a least squares fit with index @xmath72 .
the diagonal short line shows the effect of changing the scaling radius by a factor of two .
7. relation between the sfr for the normal disk and starburst samples and the ratio of the gas density to the disk orbital timescale , as described in the text .
the symbols are the same as in figure 6 .
the line is a median fit to the normal disk sample , with the slope fixed at unity as predicted by equation ( 7 ) . | measurements of , hi , and co distributions in 61 normal spiral galaxies are combined with published far - infrared and co observations of 36 infrared - selected starburst galaxies , in order to study the form of the global star formation law , over the full range of gas densities and star formation rates ( sfrs ) observed in galaxies .
the disk - averaged sfrs and gas densities for the combined sample are well represented by a schmidt law with index @xmath0 .
the schmidt law provides a surprisingly tight parametrization of the global star formation law , extending over several orders of magnitude in sfr and gas density .
an alternative formulation of the star formation law , in which the sfr is presumed to scale with the ratio of the gas density to the average orbital timescale , also fits the data very well .
both descriptions provide potentially useful recipes " for modelling the sfr in numerical simulations of galaxy formation and evolution . |
entanglement , first recognized as the characteristic trait of quantum mechanics @xcite , has been used for a long time as the main indicator of the quantumness of correlations . indeed , as shown in ref .
@xcite , for pure - state computation , exponential speed - up occurs only if entanglement grows with the size of the system .
however , the role played by entanglement in mixed - state computation is less clear .
for instance , in the so - called deterministic quantum computation with one qubit ( dqc1 ) protocol @xcite , quantum speed - up can be achieved using factorized states . as shown in ref .
@xcite , speed - up could be due to the presence of another quantifier , the so called quantum discord @xcite , which is defined as the difference between two quantum analogs of the classical mutual information .
the relationship between entanglement and quantum discord has not been completely understood , since they seem to capture different properties of the states . in ref .
@xcite , it is shown that even if quantum discord and entanglement are equal for pure states , mixed states maximizing discord in a given range of classical correlations are actually separable .
the relation between discord and entanglement has been discussed in refs .
@xcite , and an operational meaning in terms of state merging has been proposed in @xcite .
recently , the use of quantum discord has been extended to multipartite states .
a measure of genuinely multipartite quantum correlations has been introduced in @xcite . in ref .
@xcite , an attempt to generalize the definition of quantum discord in multipartite systems based on a collective measure has been proposed . in ref .
@xcite , the authors proposed different generalizations of quantum discord depending on the measurement protocol performed .
entanglement in multipartite systems has been shown to obey monogamy in the case of qubits @xcite and continuous variables @xcite .
monogamy means that if two subsystems are highly correlated , the correlation between them and other parties is bounded .
_ proved that , unlike entanglement , quantum discord is in general not monogamous @xcite .
they also suggested , based on numerical results , that w states are likely to violate monogamy .
in this brief report , using the koashi - winter formula @xcite , we prove that , for pure states , quantum discord and entanglement of formation obey the same monogamy relationship , while , for mixed states , distributed discord exceeds distributed entanglement .
then , we give an analytical proof of the violation of monogamy by all w states .
furthermore , we suggest the use of the interaction information @xcite as a measure of monogamy for the mutual information . finally , as a further application of koashi - winter equality , we prove , the conjecture on upper bounds of quantum discord and classical correlations formulated by luo _
et al . _
@xcite for rank 2 states of two qubits .
in classical information theory , mutual information between parties @xmath0 and @xmath1 is defined as @xmath2 , where @xmath3 is the shannon entropy and @xmath4 is the conditional shannon entropy of @xmath0 after @xmath1 has been measured . an equivalent formulation
[ @xmath5 can be obtained using bayes rules , because of which @xmath6 . on the other hand ,
if we try to quantize these quantities , replacing probabilities with density matrices and the shannon entropy with the von neumann entropy , their counterparts differ substantially @xcite .
the quantum mutual information is defined as @xmath7 where @xmath8 is the von neumann entropy and @xmath9 are the reduced states after tracing out party @xmath10 , while the quantized version of @xmath11 measures the classical part of the correlations @xcite and it is given by @xmath12,\label{clas}\ ] ] with the conditional entropy defined as @xmath13 , @xmath14 and where @xmath15 is the density matrix after a positive operator valued measure ( povm ) @xmath16 has been performed on @xmath1 . in some cases , orthogonal measurements are enough to find the maximum in eq .
( [ clas ] ) @xcite .
quantum discord is thus defined as the difference between @xmath17 and @xmath18 : @xmath19.\ ] ] quantum discord can be considered as a measure of how much disturbance is caused when trying to learn about party @xmath0 when measuring party @xmath1 , and has been shown to be null only for a set of states with measure zero @xcite . both classical correlations and quantum discord are asymmetric under the exchange of the two sub - parties ( i.e. , @xmath20 and @xmath21 ) .
while @xmath22 is invariant under local unitary transformations and can not increase under local operations and classical communication , @xmath23 is not monotonic under local operations .
for instance , in @xcite , it is shown how to create quantum correlations under the action of local noise .
given a measure of correlation @xmath24 , monogamy implies a tradeoff on bipartite correlations distributed along all the partitions @xmath25 ( @xmath26 ) : @xmath27 coffman , kundu , and wootters @xcite showed that this property applies to three - qubit states once the square of the concurrence @xcite ( @xmath28 ) plays the role of @xmath24 .
the extension of the proof to @xmath29-partite ( @xmath30 ) qubit systems has been given in ref .
as pointed out in @xcite , however , entanglement of formation does not satisfy the criterion given in eq .
( [ monog ] ) . then
, even if people usually refers to entanglement as a monogamous quantity , it would be worth paying attention to the entanglement monotone in use . in trying to apply this property to quantum discord , prabhu _ et al .
_ showed that monogamy is obeyed if and only if the interrogated interaction information is less than or equal to the unmeasured interaction information @xcite .
then , the authors found through numerical simulations that the subset of w states are not monogamous , in contrast with greenberger - horne - zeilinger ( ghz ) states , which can be monogamous or not . here , we prove that , for pure states , the monogamy equations for quantum discord and for entanglement of formation coincide .
let us consider the pure tripartite state @xmath31 .
quantum discord of any of the couples of sub - parties is given by @xmath32 , where @xmath33 , while @xmath34 .
as shown in ref .
@xcite , the following relationships between conditional entropies and entanglement of formation ( @xmath35 ) holds : @xmath36 this formula allows one to write @xmath37 using eq .
( [ kw ] ) , monogamy equation @xmath38 is then equivalent to @xmath39 where @xmath40 has been employed .
the equality of conservation law for distributed entanglement of formation and quantum discord , even if not associated to monogamy , was already noticed by fanchini _
et al . _
because of this equivalence , the violation of eq .
( [ entmon ] ) by w states , whose numerical evidence has been given in ref .
@xcite , admits an analytical proof .
let us recall that , apart from local operations , a generic pure state of three qubits , belonging to the ghz class , can be written as @xmath41 @xcite .
the family of w states is obtained fixing @xmath42 . as shown by coffman , kundu , and wootters @xcite , w states have zero three - tangle
; that is , they obey @xmath43 to show that eq .
( [ concw ] ) implies @xmath44 , it is enough to note that @xmath35 is a concave function of @xmath28 , since @xmath45 $ ] , where @xmath46 is the binary entropy @xmath47 , and both @xmath35 and @xmath48 admit values between @xmath49 and @xmath50 .
then , if we apply the mapping from @xmath28 to @xmath35 to the three elements of eq .
( [ concw ] ) , we find @xmath51 if @xmath52 ( i.e. for biseparable states ) and @xmath53 otherwise . as noticed in ref .
@xcite , ghz states can be monogamous or not .
actually , a numerical analysis shows that about half of them do not respect monogamy . to see a transition from observation to violation of monogamy , we consider the family of states @xmath54 .
note that @xmath55 is the maximally entangled ghz state @xmath56 , while @xmath57 coincides with the maximally entangled w state @xmath58 . for @xmath59 , qubit @xmath0 is factorized , and ( [ discmon ] ) becomes an equality . in fig .
[ figure ] , @xmath60 is plotted as a function of @xmath61 for different values of @xmath62 . as expected , for any @xmath63 , there is a threshold for @xmath61 above which the states are monogamous . , as quantified by @xmath64 , as a function of @xmath61 for different values of @xmath62 ( see the main text ) .
states are monogamous where the respective curves are positive .
black solid line is for @xmath65 , black dashed line is for @xmath66 , orange ( gray ) solid line is for @xmath67 , and orange ( gray ) dashed line is for @xmath68 .
according to the analytical proof , for w states ( @xmath65 ) @xmath60 is always positive ( i.e. , these states are never monogamous ) . for ghz states , there exists a threshold value of @xmath61 above which monogamy is satisfied .
this threshold goes to zero for vanishing @xmath61 , since in this case qubit @xmath0 becomes factorized and all the related entanglement quantifiers vanish as well.,title="fig:",width=302 ] + once the assumption of pure state is relaxed , eq .
( [ kw ] ) becomes an inequality : @xmath69 .
then , @xmath70 , or , using the subadditivity of von neumann entropy , @xmath71 thus , for mixed states , monogamy of quantum discord has a stricter bound than monogamy of entanglement .
the search for monogamy of correlations can be extended to @xmath72 .
it is actually easy to note that , for pure tripartite states , monogamies of quantum discord and classical correlations are complementary , that is @xmath73 . to prove it , it is sufficient to observe that mutual information obeys @xmath74 the generalization of eq .
( [ minfmon ] ) to mixed states presents some interesting aspects .
we have @xmath75 where @xmath76 is called correlation information @xcite . in the language of density matrices , it can be defined as @xmath77 where @xmath78 are all the possible strings containing integer numbers between @xmath50 and @xmath29 , with @xmath79 for any @xmath80 , and @xmath81 counts the length of each string .
for instance , for a bipartite system , @xmath82 coincides with the ordinary mutual information , and in the tripartite case @xmath83 .
it can be checked that , for @xmath29 odd , @xmath84 for any pure state . in classical information theory ,
the interaction information has been introduced with the aim of measuring the information that is contained in a given set of variables and that can not be accounted for considering any possible subset of them .
it should then measure genuine @xmath29-partite correlations .
actually , @xmath85 can be negative .
thus , according to the criteria given , for instance , in ref .
@xcite , it can not be used as a correlation measure .
then , its meaning is widely debated .
equation ( [ minfxmon ] ) suggests that it plays the role played by the tangle in the distribution of @xmath28 @xcite , since it is invariant under index permutation , and it can be called a `` mutual information tangle . '' when @xmath86 is negative , it quantifies the lack of monogamy of the mutual information . as shown by prabhu _
, monogamy of discord relies on the relationship between @xmath86 and its interrogated version @xcite .
@xcite , it has been conjectured that , given a bipartite state @xmath87 , defined in the hilbert space @xmath88 , the following upper bounds for quantum discord and classical correlations could exist : @xmath89 , \\{\cal j}_{a , b}&\le & \min [ s(\varrho_{a}),s(\varrho_{b})].\end{aligned}\ ] ] it is trivial to prove the existence of such an upper bound for entanglement of formation , its definition being based on the convex roof construction .
if inequality ( [ discmix ] ) were an equality , it would be easy to extend the proof to discord .
actually , inequality ( [ discmix ] ) is telling us that distributed discord could exceed distributed entanglement , and these upper bounds could be violated . while a a partial proof of the conjecture has been given in ref .
@xcite using the language of quantum operations , a full proof for the case of rank 2 states of two qubits can be given using the koashi - winter formula . by applying a purification procedure ,
we add an ancillary hilbert space @xmath90 and write a pure tripartite state @xmath91 such that @xmath92 . since @xmath87 has rank 2 , @xmath91 is a three - qubit state . as a consequence of eq .
( [ kw ] ) , we have @xmath93 then , inequalities @xmath94 and @xmath95 are immediately verified .
let us now separately discuss the cases @xmath96 and @xmath97 .
in the first case , we only need to prove @xmath98 . in ref .
@xcite , using the invariance under index permutation of the three tangle introduced by coffman , kundu , and wootters @xcite , we proved that , for the case of three qubits , if @xmath96 , then @xmath99 .
this chain rule implies @xmath100 , and then @xmath101 as we wanted to prove .
assuming now @xmath97 , we are left to show that @xmath102 .
writing explicitly @xmath103 , we use the chain rule to write @xmath104 and to obtain @xmath105 this ends the proof .
we have studied the monogamy properties of pure tripartite state .
we have shown that quantum discord and entanglement of formation obey the same monogamy relationship . applying this equivalence to the case of three qubits ,
we have shown , by analytical demonstration , that , for all the w states , quantum discord is not monogamous , in contrast with ghz states , where discord can be monogamous or not . in an example , we have shown the transition from monogamy to absence of monogamy for a subfamily of ghz states .
the equivalence between quantum discord and entanglement of formation concerning monogamy raises a subtle question that it is worth considering . while people usually claim that , for qubits , entanglement is monogamous , all we know is that there exists an entanglement monotone ( the square of the concurrence ) that is in fact monogamous . by analogy , we can say that the results of ref .
@xcite do not exhaust the search for monogamy of quantum discord and other correlations , where monogamous monotone indicators could be found . using the connection between discord end entanglement of formation
, we have also shown that in the case of rank 2 states of two qubits , as conjectured by luo _
et al . _
@xcite , quantum discord and classical correlations are bounded from above by the single - qubit von neumann entropies . a full proof can not be given because , for mixed states , the equality of conservation law for distributed entanglement of formation and quantum discord is broken .
g. adesso and f. illuminati , new j. phys . *
8 * , 15 ( 2006 ) ; g. adesso , a. serafini , and f. illuminati , phys . rev .
a * 73 * , 032345 ( 2006 ) ; t. hiroshima , g. adesso , and f. illuminati , phys . rev . lett . * 98 * , 050503 ( 2007 ) .
s. campbell , t. j. g. apollaro , c. di franco , l. banchi , a. cuccoli , r. vaia , f. plastina , and m. paternostro , arxiv:1105.5548 ; f. ciccarello and v. giovannetti , arxiv:1105.5551 ; a. streltsov , h. kampermann , and d. bruss , phys .
lett . * 107 * , 170502 ( 2011 ) .
a. acn , a. andrianov , l. costa , e. jan , j. i. latorre , and r. tarrach , phys .
lett . * 85 * , 1560 ( 2000 ) ; w. dr , g. vidal , and j. i. cirac , phys .
a * 62 * , 062314 ( 2000 ) ; a. acn , d. bruss , m. lewenstein , and a. sanpera , phys .
lett . * 87 * , 040401 ( 2001 ) . | in contrast with entanglement , as measured by concurrence , in general , quantum discord does not possess the property of monogamy , that is , there is no tradeoff between the quantum discord shared by a pair of subsystems and the quantum discord that both of them can share with a third party . here
, we show that , as far as monogamy is considered , quantum discord of pure states is equivalent to the entanglement of formation .
this result allows one to analytically prove that none of the pure three - qubit states belonging to the subclass of w states is monogamous .
a suitable physical interpretation of the meaning of the correlation information as a quantifier of monogamy for the total information is also given .
finally , we prove that , for rank 2 two - qubit states , discord and classical correlations are bounded from above by single - qubit von neumann entropies . |
in 2003 damanik and lenz @xcite completely described factors of length @xmath0 of a sturmian word which occur as @xmath1 powers for every @xmath2 and @xmath3 .
damanik and lenz prove a series of results concerning how factors of a sturmian word align to the corresponding ( finite ) standard words . by a careful analysis of the alignment
, they obtain the complete description of powers thanks to known results on powers of standard words .
our method is based on the dynamical view of sturmian words as codings of irrational rotations .
translating word - combinatorial concepts into corresponding dynamical concepts allows us to apply powerful results from diophantine approximation theory ( such as the three distance theorem ) providing a more geometric proof of the result of damanik and lenz .
our methods allow us to avoid tricky alignment arguments making the proof in our opinion easier to follow .
furthermore , the results allow us to infer a formula for the fractional index of a sturmian word based on the continued fraction expansion of its slope .
this formula and its proof appeared in an earlier paper by damanik and lenz @xcite and was also established purely combinatorially using alignment arguments .
the formula was independently obtained with different methods by carpi and de luca @xcite and justin and pirillo @xcite . for partial results and works related to powers in sturmian words
see e.g. the papers of mignosi @xcite , berstel @xcite , vandeth @xcite , and justin and pirillo @xcite .
the paper is organized as follows : in we briefly recall results concerning continued fractions and rational approximations and prove the purely number - theoretic and important for later use in . in we state needed facts about sturmian words with appropriate references .
contains the main results and their proofs .
every irrational real number @xmath4 has a unique infinite continued fraction expansion @xmath5 = a_0 + \dfrac{1}{a_1 + \dfrac{1}{a_2 + \dfrac{1}{a_3 + \cdots } } } \end{aligned}\ ] ] with @xmath6 and @xmath7 for all @xmath8 .
the numbers @xmath9 are called the _ partial quotients _ of @xmath4 .
good references on continued fractions are the books of khinchin @xcite and cassels @xcite .
we focus here only on irrational numbers , but we note that with small tweaks much of what follows also holds for rational numbers , which have finite continued fraction expansions .
the _ convergents _ @xmath10 of @xmath4 are defined by the recurrences @xmath11 the sequence @xmath12 converges to @xmath4 .
moreover , the even convergents are less than @xmath4 and form an increasing sequence and , on the other hand , the odd convergents are greater than @xmath4 and form a decreasing sequence . if @xmath13 and @xmath14 , then between the convergents @xmath15 and @xmath16 there are _ semiconvergents _ ( called intermediate fractions in khinchin s book @xcite ) which are of the form @xmath17 with @xmath18 . when the semiconvergents ( if any ) between @xmath15 and @xmath16 are ordered by the size of their denominators , the obtained sequence is increasing if @xmath19 is even and decreasing if @xmath19 is odd . note that we make a clear distinction between convergents and semiconvergents , i.e. , convergents are not a specific subtype of semiconvergents .
for the rest of this paper we make the convention that @xmath4 refers to an irrational number with a continued fraction expansion as in having convergents @xmath20 and semiconvergents @xmath21 as above .
a rational number @xmath22 is a _
best approximation _ of the real number @xmath4 if for every fraction @xmath23 such that @xmath24 and @xmath25 it holds that @xmath26 in other words , any other multiple of @xmath4 with a coefficient at most @xmath27 is further away from the nearest integer than is @xmath28 .
the next proposition shows that the best approximations of an irrational number are connected to its convergents ( for a proof see theorems 16 and 17 of @xcite ) .
[ prp : best_approximation_convergent ] the best rational approximations of an irrational number are exactly its convergents .
we identify the unit interval @xmath29 with the unit circle @xmath30 .
let @xmath31 be irrational .
the map @xmath32 where @xmath33 stands for the fractional part of the number @xmath34 , defines a rotation on @xmath30 . the circle partitions into the intervals @xmath35 and @xmath36 .
points in the same interval of the partition are said to be on the same side of @xmath37 , and points in different intervals are said to be on the opposite sides of @xmath37 .
( we are not interested in the location of the point @xmath38 . )
the points @xmath39 and @xmath40 are always on the opposite sides of @xmath37 .
the points @xmath41 with @xmath42 always lie between the points @xmath43 and @xmath39 ; see .
we measure the shortest distance to @xmath37 on @xmath30 by setting @xmath44 we have the following facts for @xmath13 and for all @xmath45 such that @xmath42 : @xmath46 we can now interpret as @xmath47 note that rotating preserves distances ; a fact we will often use without explicit mention . in particular , the distance between the points @xmath48 and @xmath49 is @xmath50 .
thus by the minimum distance between the distinct points @xmath48 and @xmath49 with @xmath51 is at least @xmath52 .
the formula tells what is the point closest to @xmath37 among the points @xmath48 for @xmath53 .
we are also interested to know the point closest to @xmath37 on the side opposite to @xmath40 .
the next result is very important and concerns this .
[ prp : closest ] let @xmath4 be an irrational number .
let @xmath0 be an integer such that @xmath54 with @xmath13 and @xmath42 .
if @xmath55 , then @xmath56 for some integer @xmath57 such that @xmath58 .
suppose that @xmath55 , and assume for a contradiction that the point @xmath48 is on the same side of @xmath37 as @xmath43 .
since @xmath59 , we conclude that @xmath60 for @xmath61 . by and
our assumption that @xmath62 , we see that @xmath60 with @xmath63 . as @xmath64 by
, we infer that the point @xmath48 must lie between the points @xmath65 and @xmath66 for some @xmath67 such that @xmath68 .
the distance between the points @xmath48 and @xmath69 is less than @xmath52 . by
, it must be that @xmath70 ; a contradiction .
suppose for a contradiction that @xmath0 is not a multiple of @xmath71 .
then the point @xmath48 lies between the points @xmath72 and @xmath73 for some @xmath74 such that @xmath75 .
as @xmath48 is on the same side of @xmath37 as the point @xmath40 , it follows that @xmath76 and @xmath77 .
the distance between the points @xmath48 and @xmath72 is less than @xmath52 , so by it must be that @xmath78 . thus necessarily @xmath79 .
using we see that the distance between the points @xmath80 and @xmath43 is @xmath81 . since @xmath82 , we infer that @xmath83 therefore by our assumption , @xmath84 so @xmath85 ; a contradiction .
we have thus concluded that @xmath56 for some @xmath86 .
let us now analyze the upper bound on @xmath57 .
first of all , @xmath87 exactly when @xmath88 .
it follows that @xmath89 .
by @xmath90 so @xmath91 .
we conclude that @xmath92 . the inequalities and imply that @xmath93 .
we derive the following useful fact : @xmath94 we need the famous three distance theorem ( see e.g. @xcite and the references therein ) .
[ thm : three_distance ] let @xmath4 be an irrational number , and let @xmath95 be a positive integer uniquely expressed in the form @xmath96 with @xmath97 and @xmath98 .
the points @xmath99 partition the circle @xmath30 into @xmath100 intervals .
there are exactly * @xmath101 intervals of length @xmath52 , * @xmath102 intervals of length @xmath103 , and * @xmath104 intervals of length @xmath105
. by the intervals of the last type ( if they exist ) are the longest , and their length is the sum of the two other length types .
we mention here only few key concepts from combinatorics on words ; good background references are lothaire s books @xcite .
a word is _ primitive _ if it is not a non - trivial power of some word .
a word @xmath106 is primitive if and only if it occurs exactly twice in @xmath107 .
the _ cyclic shift operator _
@xmath108 is defined by @xmath109 where the @xmath9 are letters .
a word @xmath110 is _ conjugate _ to @xmath111 if @xmath112 for some @xmath113 such that @xmath114 .
the _ reversal _ of the word @xmath115 where the @xmath9 are letters is the word @xmath116 .
sturmian words are a well - known class of infinite , aperiodic binary words over @xmath117 with minimal factor complexity .
they are defined as the ( right-)infinite words having @xmath100 factors of length @xmath0 for every @xmath2 . for our purposes
it is more convenient to view sturmian words equivalently as the infinite words obtained as codings of orbits of points in an irrational circle rotation with two intervals @xcite .
let us make this more precise .
the frequency @xmath4 of letter @xmath118 ( called the _ slope _ ) in a sturmian words exists , and it is irrational . divide the circle @xmath30 into two intervals @xmath119 and @xmath120 defined by the points @xmath37 and @xmath121 , and define the coding function @xmath122 by setting @xmath123 if @xmath124 and @xmath125 if @xmath126 .
the coding of the orbit of a point @xmath34 is the infinite word @xmath127 obtained by setting its @xmath128 letter to equal @xmath129 where @xmath130 is the rotation by angle @xmath4 .
this word is sturmian with slope @xmath4 , and conversely every sturmian word with slope @xmath4 is obtained this way . to make the definition proper , we need to define how @xmath122 behaves in the endpoints @xmath37 and @xmath121 .
we have two options : either take @xmath131 and @xmath132 or @xmath133 $ ] and @xmath134 $ ] .
the difference is seen in the codings of the orbits of the special points @xmath135 , and both options are needed to be able to obtain every sturmian word of slope @xmath4 as a coding of a rotation .
however , in this paper we are not concerned about this choice .
we make the convention that @xmath136 with @xmath137 is either of the half - open intervals of @xmath30 separated by the points @xmath34 and @xmath138 ( taken modulo @xmath118 if necessary ) not containing the point @xmath37 as an interior point .
the interval @xmath139 is either of the half - open intervals separated by the points @xmath37 and @xmath34 having smallest length ( the case @xmath140 is not important in this paper ) . since the sequence @xmath141 is dense in @xmath29as is well - known every sturmian word of slope @xmath4 has the same language ( that is , the set of factors ) ; this language is denoted by @xmath142 . thus to study repetitions , it is sufficient to analyze @xmath142 .
the _ fractional index _
@xmath143 of a nonempty factor @xmath144 is defined as @xmath145 where the _ fractional power _
@xmath146 is the word @xmath147 with @xmath148 and @xmath149 .
the _ index _
@xmath150 of a nonempty factor @xmath106 is defined similarly by letting @xmath19 take only integral values .
the index of a factor in @xmath142 is always finite .
the _ fractional index of a sturmian word _ with slope @xmath4 is defined to be @xmath151 this quantity can be infinite . for every factor @xmath152 of length @xmath0 there exists a unique subinterval @xmath153 $ ] of @xmath30 such that @xmath127 begins with @xmath106 if and only if @xmath154 $ ] .
clearly @xmath155 = i_{a_0 } \cap r^{-1}(i_{a_1 } ) \cap \ldots \cap r^{-(n-1)}(i_{a_{n-1 } } ) .
\end{aligned}\ ] ] we denote the length of the interval @xmath153 $ ] by @xmath156|$ ] .
the points @xmath157 partition the circle into @xmath100 intervals , which have one - to - one correspondence with the words of @xmath142 of length @xmath0 . among these intervals
the interval containing the point @xmath158 corresponds to the right special factor of length @xmath0 .
a factor @xmath106 is _ right special _ if both @xmath159 .
similarly a factor is _ left special _ if both @xmath160 . in a sturmian word
there exists a unique right special and a unique left special factor of length @xmath0 for all @xmath2 .
the language @xmath142 is mirror - invariant , that is , for every @xmath144 also @xmath161 .
it follows that the right special factor of length @xmath0 is the reversal of the left special factor of length @xmath0 .
given the continued fraction expansion of an irrational @xmath31 , we define the corresponding _ standard sequence _
@xmath162 of words by @xmath163 as @xmath164 is a prefix of @xmath165 for @xmath8 , the sequence @xmath166 converges to a unique infinite word @xmath167 called the infinite standard sturmian word of slope @xmath4 , and it equals @xmath168 .
inspired by the notion of semiconvergents , we define _ semistandard _ words for @xmath13 by @xmath169 with @xmath18 .
clearly @xmath170 and @xmath171 .
every prefix of @xmath167 is left special , so in particular , standard and semistandard words are left special .
every standard or semistandard word is primitive ( * ? ? ?
* proposition 2.2.3 ) .
an important property of standard words is that the words @xmath164 and @xmath172 almost commute ; namely @xmath173 and @xmath174 for some word @xmath106 and distinct letters @xmath175 and @xmath27 . for more information about standard words see chapter 2 of @xcite and berstel s paper @xcite .
here we see that the only difference between the words @xmath167 and @xmath176 where @xmath177 $ ] and @xmath178 $ ] is that the roles of the letters @xmath37 and @xmath118 are reversed .
thus for the study of powers , we may assume without loss of generality that @xmath179 .
for the rest of this paper we make the convention that the partial quotients of an irrational @xmath4 satisfy @xmath180 and @xmath179 , that is , @xmath181 .
moreover , the words @xmath164 and @xmath182 refer to the standard or semistandard words of slope @xmath4 .
this section presents a complete description of powers occurring in a sturmian word with slope @xmath4 . as a side - product , in we obtain a description of conjugacy classes of length @xmath183 .
finally , as an easy consequence of the established results , we obtain a formula for the fractional index of a sturmian word ( ) .
the following important proposition shows the usefulness of in the study of sturmian words and plays a role similar to theorem 1 of @xcite .
[ prp : square_length ] if @xmath184 with @xmath106 primitive , then @xmath185 for some @xmath186 or @xmath187 for some @xmath13 with @xmath188 .
let @xmath189 .
if @xmath190 , then the factors of length @xmath0 are readily seen to be @xmath191 and the conjugates of @xmath192 .
since the minimum number of letters @xmath37 between two occurrences of letter @xmath118 in @xmath142 is @xmath193 and the maximum number is @xmath194 , the only way @xmath107 can be a factor is that @xmath195 .
suppose then that @xmath196 and @xmath153 = i(-i\alpha , -j\alpha)$ ] with @xmath197 .
we may assume without loss of generality that @xmath106 is right special , so @xmath198 $ ] .
further , since @xmath199 = [ w ] \cap r^{-n}([w ] ) \neq \emptyset$ ] , then necessarily ( depending on @xmath0 ) either @xmath199 = i(-i\alpha , -(j+n)\alpha)$ ] or @xmath199 = i(-j\alpha , -(i+n)\alpha)$ ] .
we assume that @xmath199 = i(-i\alpha , -(j+n)\alpha)$ ] ; the other case is symmetric .
we wish to prove that the points @xmath200 and @xmath201 are actually the same point .
this is equivalent to saying that @xmath202 .
assume on the contrary that @xmath203 .
let @xmath175 be the first letter of @xmath106 and @xmath27 be a letter such that @xmath204 .
note that @xmath199 \subset [ wa]$ ] .
now as @xmath106 is right special , @xmath205 = i(-j\alpha , -(n+1)\alpha)$ ] and @xmath206 = i(-(n+1)\alpha , -i\alpha)$ ] .
let @xmath207 $ ] and @xmath208\setminus[w^2]$ ] , and let @xmath110 be the longest common prefix of @xmath127 and @xmath209 . since @xmath199 \neq [ wa]$ ] , we have that @xmath210 .
moreover , @xmath110 is right special , so @xmath106 is a suffix of @xmath110 .
however , @xmath107 is a prefix of @xmath127 implying that @xmath110 is a prefix of @xmath107 .
thus @xmath107 contains at least three occurrences of @xmath106 contradicting the primitivity of @xmath106 . from this contradiction
we conclude that @xmath202 .
there are no points @xmath211 in the interval @xmath212 with @xmath213 .
therefore the point @xmath135 is the closest point to @xmath37 from either side .
if @xmath214 , then it must be that @xmath215
. otherwise let @xmath13 be such that @xmath216 with @xmath42 . by either @xmath217 or @xmath218 proving the claim .
indeed , for each length given in the statement of the previous proposition , there exists a factor occurring as a square .
[ lem : standard_square ] we have that @xmath219 and @xmath220 for all @xmath13 and @xmath45 such that @xmath42 . as @xmath221 and @xmath222 , clearly @xmath219 .
since the words @xmath223 and @xmath224 differ only by their last two letters , it follows that @xmath225 is a prefix of @xmath223 if @xmath13 . as @xmath164 is a prefix of @xmath165 when @xmath186 , the word @xmath226 is both a prefix and a suffix of @xmath227 for all @xmath13 and @xmath45 such that @xmath42 .
thus @xmath225 contains @xmath228 .
the claim follows .
as was seen in the proof of , the index of a factor @xmath106 of length @xmath0 depends only on the maximum @xmath229 such that @xmath230 $ ] for @xmath231 where @xmath34 is either of the endpoints of @xmath153 $ ] .
that is , the index of a factor depends only on the length of its interval but not on its position . to put it more precisely , if @xmath106 is a factor of length @xmath0 , then @xmath232|}{\||w|\alpha\| } \right\rfloor \end{aligned}\ ] ] where @xmath233 is @xmath118 if @xmath156| \neq \||w|\alpha\|$ ] and @xmath37 otherwise .
next we will carefully characterize the lengths of the intervals of factors of length @xmath183 .
after this it is easy to conclude the main results .
[ thm : conjugate_interval_lengths ] let @xmath218 with @xmath13 and @xmath42 .
then @xmath234 for @xmath235 .
the intervals of the first @xmath236 conjugates of @xmath237 have length @xmath105 , and the intervals of the latter @xmath238 conjugates have length @xmath52 .
the interval of the remaining factor has length @xmath103 .
the geometric ideas of this proof are illustrated in the example following this proof .
the intervals of the factors of length @xmath0 are called in this proof _
level @xmath0 intervals_. with the same effort we prove here more than what is claimed above ; we give the exact positions of the intervals of the conjugates of @xmath237 on the circle . by the interval
@xmath239 has exactly one point @xmath240 with @xmath241 as an interior point ; namely the point @xmath135 .
that is , the point @xmath135 split the level @xmath242 interval @xmath243 into the level @xmath0 intervals @xmath244 and @xmath245 .
observe that @xmath246 .
the three distance theorem tells that level @xmath0 intervals have lengths @xmath247 and @xmath52 .
in particular , interval @xmath248 is the unique level @xmath0 interval of length @xmath103 .
let @xmath113 be the smallest positive integer such that the interval @xmath249 is not any interval of level @xmath0 .
the interval @xmath249 must be a union of two level @xmath0 intervals : one having length @xmath250 and the other having length @xmath251 .
this is true as by it can be deduced that @xmath252 is never a multiple of @xmath250 ; further , @xmath250 is never a multiple of @xmath251 .
since the interval of length @xmath103 is unique , we conclude that the other interval in the union is @xmath248 . as @xmath4 is irrational , @xmath253 , so it must be that @xmath254 where @xmath255 . therefore @xmath256 maps the endpoint @xmath37 of @xmath248 to the endpoint @xmath257 of @xmath258 , so @xmath259 . as @xmath260 , also @xmath261 .
we have shown that the level @xmath0 intervals @xmath262 have length @xmath263 . by the three distance theorem the remaining @xmath101 intervals excluding @xmath248 have length @xmath52 .
what remains is to analyze the connection between rotation and conjugation .
let @xmath110 and @xmath111 be factors of length @xmath0 such that @xmath264 = m$ ] and @xmath265 = l$ ] . since the intervals @xmath258 and @xmath248 are on the opposite sides of @xmath37 , we have that @xmath266 and @xmath267 for distinct letters @xmath175 and @xmath27 .
let @xmath268 and @xmath269 .
since @xmath261 , the interval @xmath270 is the interval of some factor @xmath106 of length @xmath0 . therefore the sturmian words @xmath271 and @xmath272 both have @xmath106 as a prefix .
thus @xmath127 begins with @xmath273 and @xmath209 begins with @xmath274 .
hence @xmath106 must be left special , that is , @xmath275 .
we will show next that @xmath111 is not conjugate to @xmath182 .
note that @xmath276 is odd if and only if @xmath277 .
hence the first letter of @xmath111 is @xmath37 if and only if @xmath276 is odd . on the other hand ,
the last letter of @xmath182 is @xmath37 if and only if @xmath276 is even .
thus we conclude that the first letter of @xmath111 is distinct from the last letter of @xmath182 .
however , as the suffix of @xmath111 of length @xmath242 is a prefix of @xmath182 , we see that there are more letters @xmath27 in @xmath111 than there are in @xmath182 , so @xmath111 and @xmath182 can not be conjugate .
let then @xmath278 be the factor of length @xmath0 such that @xmath279 = r^{-1}(j)$ ] . since @xmath280 , it must be that @xmath281 $ ] .
thus @xmath278 is right special , that is , @xmath282 . by @xmath220 .
thus every conjugate of @xmath182 is a factor .
further , by the mirror - invariance of @xmath142 , we see that @xmath182 and @xmath237 are conjugates .
moreover , every conjugate of @xmath237 is extended to the left by its last letter .
suppose that @xmath283 is a factor of length @xmath0 such that @xmath284)$ ] is the interval of some factor @xmath285 of length @xmath0 .
as @xmath286)$ ] does not satisfy this condition , it follows that @xmath287 , so @xmath288 extends to the left uniquely .
we will prove that @xmath289 .
write @xmath290 for some letter @xmath291 .
then obviously @xmath292 for some letter @xmath293 . by definition @xmath285
must be followed by the letter @xmath291 , that is , @xmath294 .
we have that @xmath295 because @xmath288 is uniquely extended to the left by its last letter .
therefore we conclude that @xmath289 . in this way
we see that the factors of length @xmath0 having the intervals @xmath262 correspond ( in order ) to the factors @xmath296 .
we saw above that @xmath111 is not conjugate to @xmath182 , so it must be that @xmath297 .
thus the factors of length @xmath0 having the intervals @xmath264 = l , r^{-1}(l ) , r^{-2}(l ) , \ldots , r^{-(n - q_{k-1})}(l)$ ] correspond ( in order ) to the factors @xmath298 .
as @xmath299 , we have a complete description of the positions of the intervals of conjugates of @xmath237 using the backward orbit of @xmath243 under @xmath130 .
= [ line width=0.8pt ] = [ line width=2.0pt , black ] = [ line width=0.8pt , bluish , text = bluish , font= * * ] = [ line width=0.8pt , greenish , text = greenish , font= * * ] = [ line width=0.8pt , cyanish , text = cyanish , font= * * ] ( 0.0:2.0 ) arc ( 0.0:61.1:2.0 ) ; ( 61.1:2.0 ) arc ( 61.1:96.5:2.0 ) ; ( 96.5:2.0 ) arc ( 96.5:192.9:2.0 ) ; ( 192.9:2.0 ) arc ( 192.9:228.2:2.0 ) ; ( 228.2:2.0 ) arc ( 228.2:324.7:2.0 ) ; ( 324.7:2.0 ) arc ( 324.7:360:2.0 ) ; ( 0:2.0 ) circle(0.9pt ) ; ( 61.1:2.0 ) circle(0.9pt ) ; ( 96.5:2.0 ) circle(0.9pt ) ; ( 228.2:2.0 ) circle(0.9pt ) ; ( 192.9:2.0 ) circle(0.9pt ) ; ( 324.7:2.0 ) circle(0.9pt ) ; ( 0:2.8 ) arc ( 0:96.5:2.8 ) ; ( -35.3:3.3 ) arc ( -35.3:61.1:3.3 ) ; ( 0:2.8 ) circle(0.9pt ) ; ( 96.5:2.8 ) circle(0.9pt ) ; ( -35.3:3.3 ) circle(0.9pt ) ; ( 61.1:3.3 ) circle(0.9pt ) ; at ( 2.2,0 ) ; at ( -1.6,-1.65 ) ; at ( 1.2,2.05 ) ; at ( 1.1,0.85 ) ; at ( 0.35,1.65 ) ; at ( -1.0,1.1 ) ; at ( -1.25,-1.0 ) ; at ( 0.1,-1.75 ) ; at ( 1.45,-0.6 ) ; ( 0.1,-1.6 ) ( -1.1,0.9 ) ; ( -0.9,0.9 ) ( 1.05,-0.6 ) ; ( 1.05,-0.75 ) ( -0.85,-1.0 ) ; ( -0.9,-0.85 ) ( 0.35,1.45 ) ; at ( -1.05,2.8 ) ; at ( 1.0,3.0 ) ; at ( -1.85,1.2 ) ; at ( 0.0,-2.25 ) ; at ( 0.4,2.15 ) ; at ( -1.85,-1.2 ) ; at ( 2.1,-0.6 ) ; at ( 1.75,1.3 ) ; * * * * * * let @xmath300 = \frac{1}{2}(\sqrt{3 } - 1)$ ] ( that is , the continued fraction expansion of @xmath4 has period @xmath301 ) .
consider the semiconvergent @xmath302 of @xmath4 and factors of length @xmath303 .
the factors of length @xmath0 are @xmath304 and @xmath305 .
their intervals are depicted in .
there are intervals of type @xmath306 and @xmath307 depending on their length .
intervals of type @xmath118 have length @xmath308 , intervals of type @xmath309 have length @xmath310 and the interval of type @xmath307 has length @xmath311 . as in the proof of , the point @xmath135 has split the type @xmath118 interval @xmath312 into intervals of type @xmath309 and @xmath307 .
the interval @xmath313 corresponds to the right special factor @xmath314 .
the arrows in the figure indicate how conjugation acts on @xmath315 .
the backward orbit of @xmath243 corresponds to conjugates of @xmath315 of type @xmath118 until the interval of the left special factor @xmath316 is encountered .
as seen in the proof of , the interval @xmath317)$ ] no longer coincides with any interval of length @xmath0 . here
@xmath317 ) = l \cup m = i(0 , -n\alpha ) \cup i(0 , -3\alpha)$ ] just as the proof requires .
the factor having interval @xmath248 is here seen to be not conjugate to @xmath315 as it should be by the proof .
the interval @xmath258 must then correspond to the conjugate of @xmath316 . as in the proof ,
the rest of the conjugates of @xmath315 are obtained by rotating @xmath258 backwards .
the intervals obtained this way are of type @xmath309 .
we are now ready to prove the main result .
the result was originally proven by @xcite @xcite .
we present it here phrased in a different way .
[ thm : power_characterization ] consider indices of factors of length @xmath318 in @xmath142 , and let @xmath13 .
a. if @xmath319 , then the index of the conjugates of @xmath192 is @xmath118 , and the index of the remaining factor @xmath191 is @xmath320 .
b. if @xmath215 , then the index of the conjugates of @xmath321 is @xmath322 , and the index of the remaining factor @xmath323 is @xmath118 .
c. if @xmath324 , then the index of any of the first @xmath236 conjugates of @xmath325 is @xmath326 , the index of any of the remaining @xmath238 conjugates is @xmath327 and the index of the remaining factor is @xmath118 .
d. if @xmath218 with @xmath188 , then the index of the first @xmath236 conjugates of @xmath237 is @xmath309 , and the index of the remaining factors is @xmath118 .
e. if @xmath328 with @xmath329 , then the index of any of the first @xmath330 conjugates of @xmath331 is @xmath332 , and the index of any remaining factor is @xmath118 .
f. if @xmath333 with @xmath334 , then the index of any of the first @xmath236 conjugates of @xmath335 is @xmath336 , the index of any of the next @xmath337 conjugates is @xmath338 , and the index of any remaining factor is @xmath118 .
g. if @xmath0 does not fall into any of the above cases , then the index of every factor of length @xmath0 is @xmath118 .
first of all , observe that all the cases _ ( i)(vii ) _ are mutually exclusive .
consider the cases _ ( i ) _ and _ ( ii)_. the factors of length @xmath339 are readily seen to be @xmath191 and the conjugates of @xmath192 . as the index of @xmath37 is @xmath194 , the index of the factor
@xmath191 is @xmath320 .
the intervals of the conjugates of @xmath192 have length @xmath4 . if @xmath340 , then the index of the factor @xmath341 is @xmath118 . if @xmath342 , the number @xmath343 equals @xmath118 unless @xmath215 when it equals @xmath322 by .
the claims of _ ( i ) _ and _ ( ii ) _ follow from .
consider next a factor @xmath106 of length @xmath218 for some @xmath13 and @xmath45 such that @xmath42 . by , the intervals of the first @xmath236 conjugates of @xmath237 have length @xmath105 .
using and we see that their index equals to @xmath344 if @xmath345 , then by @xmath346 , so the index is @xmath309 .
if @xmath347 , then by , the index equals to @xmath348 .
this proves the first claims in _
( iii ) _ and _ ( iv)_. the latter cases are analogous , so _
( iii ) _ and _ ( iv ) _ are proved .
shows that the factors not covered by the cases _ ( i)(iv ) _ having index higher than @xmath118 must be nonprimitive . by _ ( i ) _ and _ ( iv ) _ they must have length @xmath349 for some @xmath8 , meaning that we are in either of the cases _ ( v ) _ or _
( vi)_. it is a straightforward application of _ ( ii ) _ and _ ( iii ) _ to deduce _ ( v ) _ and _ ( vi)_. the theorem is proved .
in particular , every sturmian word contains infinitely many cubes , but fourth powers are avoidable .
the theorem implies the following weaker version which is still useful ( compare to ( * ? ? ?
* lemma 3.6 ) ) : let @xmath144 be primitive .
if @xmath184 , then @xmath106 is conjugate to @xmath164 for some @xmath186 or to @xmath182 with @xmath13 and @xmath188 . if @xmath350 , then either @xmath351 and @xmath352 or @xmath106 is conjugate to some @xmath164 with @xmath8 .
we obtain the result of @xcite , @xcite , and @xcite on the fractional index of sturmian words as a direct consequence of the results so far : [ thm : index_formula ] the fractional index of a sturmian word with slope @xmath4 is @xmath353 the largest fractional power of a factor with length less than @xmath330 is clearly @xmath323 . therefore by it is sufficient to analyze the largest fractional power of a ( primitive ) factor of length @xmath354 for @xmath8 . by the index @xmath355 of the first @xmath236 conjugates of @xmath325
dominates the index of the rest of the factors of length @xmath354 .
the fractional part of the fractional index of a factor @xmath106 is determined by the shortest extension of @xmath106 to a right special factor .
note that from the proof of it is evident that @xmath356 .
thus among the first @xmath236 conjugates of @xmath325 , the factor @xmath164 has longest extension to a right special factor , and the length of the extension is @xmath357 .
thus the fractional index of @xmath164 is @xmath358 .
the claim follows .
in particular , this theorem says that that a sturmian word has bounded fractional index if and only if the partial quotients of its slope are bounded .
this is a result of mignosi @xcite .
an alternative proof was given by berstel @xcite .
the author was supported by university of turku graduate school utugs matti programme . | we present a new , dynamical way to study powers ( that is , repetitions ) in sturmian words based on results from diophantine approximation theory . as a result
, we provide an alternative and shorter proof of a result by damanik and lenz characterizing powers in sturmian words [ powers in sturmian sequences , eur .
j. combin . 24
( 2003 ) , 377390 ] .
further , as a consequence , we obtain a previously known formula for the fractional index of a sturmian word based on the continued fraction expansion of its slope . : sturmian word , standard word , power , combinatorics on words , continued fraction |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.